ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/libev/ev.pod

Comparing libev/ev.pod (file contents):
Revision 1.27 by root, Wed Nov 14 05:02:07 2007 UTC vs.
Revision 1.453 by root, Tue Jun 25 05:01:22 2019 UTC

		1	=encoding utf-8
		2
1	=head1 NAME	3	=head1 NAME
2		4
3	libev - a high performance full-featured event loop written in C	5	libev - a high performance full-featured event loop written in C
4		6
5	=head1 SYNOPSIS	7	=head1 SYNOPSIS
6		8
7	#include <ev.h>	9	#include <ev.h>
8		10
9	=head1 DESCRIPTION	11	=head2 EXAMPLE PROGRAM
		12
		13	// a single header file is required
		14	#include <ev.h>
		15
		16	#include <stdio.h> // for puts
		17
		18	// every watcher type has its own typedef'd struct
		19	// with the name ev_TYPE
		20	ev_io stdin_watcher;
		21	ev_timer timeout_watcher;
		22
		23	// all watcher callbacks have a similar signature
		24	// this callback is called when data is readable on stdin
		25	static void
		26	stdin_cb (EV_P_ ev_io *w, int revents)
		27	{
		28	puts ("stdin ready");
		29	// for one-shot events, one must manually stop the watcher
		30	// with its corresponding stop function.
		31	ev_io_stop (EV_A_ w);
		32
		33	// this causes all nested ev_run's to stop iterating
		34	ev_break (EV_A_ EVBREAK_ALL);
		35	}
		36
		37	// another callback, this time for a time-out
		38	static void
		39	timeout_cb (EV_P_ ev_timer *w, int revents)
		40	{
		41	puts ("timeout");
		42	// this causes the innermost ev_run to stop iterating
		43	ev_break (EV_A_ EVBREAK_ONE);
		44	}
		45
		46	int
		47	main (void)
		48	{
		49	// use the default event loop unless you have special needs
		50	struct ev_loop *loop = EV_DEFAULT;
		51
		52	// initialise an io watcher, then start it
		53	// this one will watch for stdin to become readable
		54	ev_io_init (&stdin_watcher, stdin_cb, /*STDIN_FILENO*/ 0, EV_READ);
		55	ev_io_start (loop, &stdin_watcher);
		56
		57	// initialise a timer watcher, then start it
		58	// simple non-repeating 5.5 second timeout
		59	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		60	ev_timer_start (loop, &timeout_watcher);
		61
		62	// now wait for events to arrive
		63	ev_run (loop, 0);
		64
		65	// break was called, so exit
		66	return 0;
		67	}
		68
		69	=head1 ABOUT THIS DOCUMENT
		70
		71	This document documents the libev software package.
		72
		73	The newest version of this document is also available as an html-formatted
		74	web page you might find easier to navigate when reading it for the first
		75	time: L<http://pod.tst.eu/http://cvs.schmorp.de/libev/ev.pod>.
		76
		77	While this document tries to be as complete as possible in documenting
		78	libev, its usage and the rationale behind its design, it is not a tutorial
		79	on event-based programming, nor will it introduce event-based programming
		80	with libev.
		81
		82	Familiarity with event based programming techniques in general is assumed
		83	throughout this document.
		84
		85	=head1 WHAT TO READ WHEN IN A HURRY
		86
		87	This manual tries to be very detailed, but unfortunately, this also makes
		88	it very long. If you just want to know the basics of libev, I suggest
		89	reading L</ANATOMY OF A WATCHER>, then the L</EXAMPLE PROGRAM> above and
		90	look up the missing functions in L</GLOBAL FUNCTIONS> and the C<ev_io> and
		91	C<ev_timer> sections in L</WATCHER TYPES>.
		92
		93	=head1 ABOUT LIBEV
10		94
11	Libev is an event loop: you register interest in certain events (such as a	95	Libev is an event loop: you register interest in certain events (such as a
12	file descriptor being readable or a timeout occuring), and it will manage	96	file descriptor being readable or a timeout occurring), and it will manage
13	these event sources and provide your program with events.	97	these event sources and provide your program with events.
14		98
15	To do this, it must take more or less complete control over your process	99	To do this, it must take more or less complete control over your process
16	(or thread) by executing the I<event loop> handler, and will then	100	(or thread) by executing the I<event loop> handler, and will then
17	communicate events via a callback mechanism.	101	communicate events via a callback mechanism.
…		…
19	You register interest in certain events by registering so-called I<event	103	You register interest in certain events by registering so-called I<event
20	watchers>, which are relatively small C structures you initialise with the	104	watchers>, which are relatively small C structures you initialise with the
21	details of the event, and then hand it over to libev by I<starting> the	105	details of the event, and then hand it over to libev by I<starting> the
22	watcher.	106	watcher.
23		107
24	=head1 FEATURES	108	=head2 FEATURES
25		109
26	Libev supports select, poll, the linux-specific epoll and the bsd-specific	110	Libev supports C<select>, C<poll>, the Linux-specific aio and C<epoll>
27	kqueue mechanisms for file descriptor events, relative timers, absolute	111	interfaces, the BSD-specific C<kqueue> and the Solaris-specific event port
28	timers with customised rescheduling, signal events, process status change	112	mechanisms for file descriptor events (C<ev_io>), the Linux C<inotify>
		113	interface (for C<ev_stat>), Linux eventfd/signalfd (for faster and cleaner
		114	inter-thread wakeup (C<ev_async>)/signal handling (C<ev_signal>)) relative
		115	timers (C<ev_timer>), absolute timers with customised rescheduling
		116	(C<ev_periodic>), synchronous signals (C<ev_signal>), process status
29	events (related to SIGCHLD), and event watchers dealing with the event	117	change events (C<ev_child>), and event watchers dealing with the event
30	loop mechanism itself (idle, prepare and check watchers). It also is quite	118	loop mechanism itself (C<ev_idle>, C<ev_embed>, C<ev_prepare> and
		119	C<ev_check> watchers) as well as file watchers (C<ev_stat>) and even
		120	limited support for fork events (C<ev_fork>).
		121
		122	It also is quite fast (see this
31	fast (see this L<benchmark|http://libev.schmorp.de/bench.html> comparing	123	L<benchmark|http://libev.schmorp.de/bench.html> comparing it to libevent
32	it to libevent for example).	124	for example).
33		125
34	=head1 CONVENTIONS	126	=head2 CONVENTIONS
35		127
36	Libev is very configurable. In this manual the default configuration	128	Libev is very configurable. In this manual the default (and most common)
37	will be described, which supports multiple event loops. For more info	129	configuration will be described, which supports multiple event loops. For
38	about various configuration options please have a look at the file	130	more info about various configuration options please have a look at
39	F<README.embed> in the libev distribution. If libev was configured without	131	B<EMBED> section in this manual. If libev was configured without support
40	support for multiple event loops, then all functions taking an initial	132	for multiple event loops, then all functions taking an initial argument of
41	argument of name C<loop> (which is always of type C<struct ev_loop *>)	133	name C<loop> (which is always of type C<struct ev_loop *>) will not have
42	will not have this argument.	134	this argument.
43		135
44	=head1 TIME REPRESENTATION	136	=head2 TIME REPRESENTATION
45		137
46	Libev represents time as a single floating point number, representing the	138	Libev represents time as a single floating point number, representing
47	(fractional) number of seconds since the (POSIX) epoch (somewhere near	139	the (fractional) number of seconds since the (POSIX) epoch (in practice
48	the beginning of 1970, details are complicated, don't ask). This type is	140	somewhere near the beginning of 1970, details are complicated, don't
49	called C<ev_tstamp>, which is what you should use too. It usually aliases	141	ask). This type is called C<ev_tstamp>, which is what you should use
50	to the double type in C.	142	too. It usually aliases to the C<double> type in C. When you need to do
		143	any calculations on it, you should treat it as some floating point value.
		144
		145	Unlike the name component C<stamp> might indicate, it is also used for
		146	time differences (e.g. delays) throughout libev.
		147
		148	=head1 ERROR HANDLING
		149
		150	Libev knows three classes of errors: operating system errors, usage errors
		151	and internal errors (bugs).
		152
		153	When libev catches an operating system error it cannot handle (for example
		154	a system call indicating a condition libev cannot fix), it calls the callback
		155	set via C<ev_set_syserr_cb>, which is supposed to fix the problem or
		156	abort. The default is to print a diagnostic message and to call C<abort
		157	()>.
		158
		159	When libev detects a usage error such as a negative timer interval, then
		160	it will print a diagnostic message and abort (via the C<assert> mechanism,
		161	so C<NDEBUG> will disable this checking): these are programming errors in
		162	the libev caller and need to be fixed there.
		163
		164	Libev also has a few internal error-checking C<assert>ions, and also has
		165	extensive consistency checking code. These do not trigger under normal
		166	circumstances, as they indicate either a bug in libev or worse.
		167
51		168
52	=head1 GLOBAL FUNCTIONS	169	=head1 GLOBAL FUNCTIONS
53		170
54	These functions can be called anytime, even before initialising the	171	These functions can be called anytime, even before initialising the
55	library in any way.	172	library in any way.
…		…
58		175
59	=item ev_tstamp ev_time ()	176	=item ev_tstamp ev_time ()
60		177
61	Returns the current time as libev would use it. Please note that the	178	Returns the current time as libev would use it. Please note that the
62	C<ev_now> function is usually faster and also often returns the timestamp	179	C<ev_now> function is usually faster and also often returns the timestamp
63	you actually want to know.	180	you actually want to know. Also interesting is the combination of
		181	C<ev_now_update> and C<ev_now>.
		182
		183	=item ev_sleep (ev_tstamp interval)
		184
		185	Sleep for the given interval: The current thread will be blocked
		186	until either it is interrupted or the given time interval has
		187	passed (approximately - it might return a bit earlier even if not
		188	interrupted). Returns immediately if C<< interval <= 0 >>.
		189
		190	Basically this is a sub-second-resolution C<sleep ()>.
		191
		192	The range of the C<interval> is limited - libev only guarantees to work
		193	with sleep times of up to one day (C<< interval <= 86400 >>).
64		194
65	=item int ev_version_major ()	195	=item int ev_version_major ()
66		196
67	=item int ev_version_minor ()	197	=item int ev_version_minor ()
68		198
69	You can find out the major and minor version numbers of the library	199	You can find out the major and minor ABI version numbers of the library
70	you linked against by calling the functions C<ev_version_major> and	200	you linked against by calling the functions C<ev_version_major> and
71	C<ev_version_minor>. If you want, you can compare against the global	201	C<ev_version_minor>. If you want, you can compare against the global
72	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the	202	symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the
73	version of the library your program was compiled against.	203	version of the library your program was compiled against.
74		204
		205	These version numbers refer to the ABI version of the library, not the
		206	release version.
		207
75	Usually, it's a good idea to terminate if the major versions mismatch,	208	Usually, it's a good idea to terminate if the major versions mismatch,
76	as this indicates an incompatible change. Minor versions are usually	209	as this indicates an incompatible change. Minor versions are usually
77	compatible to older versions, so a larger minor version alone is usually	210	compatible to older versions, so a larger minor version alone is usually
78	not a problem.	211	not a problem.
79		212
		213	Example: Make sure we haven't accidentally been linked against the wrong
		214	version (note, however, that this will not detect other ABI mismatches,
		215	such as LFS or reentrancy).
		216
		217	assert (("libev version mismatch",
		218	ev_version_major () == EV_VERSION_MAJOR
		219	&& ev_version_minor () >= EV_VERSION_MINOR));
		220
		221	=item unsigned int ev_supported_backends ()
		222
		223	Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*>
		224	value) compiled into this binary of libev (independent of their
		225	availability on the system you are running on). See C<ev_default_loop> for
		226	a description of the set values.
		227
		228	Example: make sure we have the epoll method, because yeah this is cool and
		229	a must have and can we have a torrent of it please!!!11
		230
		231	assert (("sorry, no epoll, no sex",
		232	ev_supported_backends () & EVBACKEND_EPOLL));
		233
		234	=item unsigned int ev_recommended_backends ()
		235
		236	Return the set of all backends compiled into this binary of libev and
		237	also recommended for this platform, meaning it will work for most file
		238	descriptor types. This set is often smaller than the one returned by
		239	C<ev_supported_backends>, as for example kqueue is broken on most BSDs
		240	and will not be auto-detected unless you explicitly request it (assuming
		241	you know what you are doing). This is the set of backends that libev will
		242	probe for if you specify no backends explicitly.
		243
		244	=item unsigned int ev_embeddable_backends ()
		245
		246	Returns the set of backends that are embeddable in other event loops. This
		247	value is platform-specific but can include backends not available on the
		248	current system. To find which embeddable backends might be supported on
		249	the current system, you would need to look at C<ev_embeddable_backends ()
		250	& ev_supported_backends ()>, likewise for recommended ones.
		251
		252	See the description of C<ev_embed> watchers for more info.
		253
80	=item ev_set_allocator (void *(*cb)(void *ptr, long size))	254	=item ev_set_allocator (void *(*cb)(void *ptr, long size) throw ())
81		255
82	Sets the allocation function to use (the prototype is similar to the	256	Sets the allocation function to use (the prototype is similar - the
83	realloc C function, the semantics are identical). It is used to allocate	257	semantics are identical to the C<realloc> C89/SuS/POSIX function). It is
84	and free memory (no surprises here). If it returns zero when memory	258	used to allocate and free memory (no surprises here). If it returns zero
85	needs to be allocated, the library might abort or take some potentially	259	when memory needs to be allocated (C<size != 0>), the library might abort
86	destructive action. The default is your system realloc function.	260	or take some potentially destructive action.
		261
		262	Since some systems (at least OpenBSD and Darwin) fail to implement
		263	correct C<realloc> semantics, libev will use a wrapper around the system
		264	C<realloc> and C<free> functions by default.
87		265
88	You could override this function in high-availability programs to, say,	266	You could override this function in high-availability programs to, say,
89	free some memory if it cannot allocate memory, to use a special allocator,	267	free some memory if it cannot allocate memory, to use a special allocator,
90	or even to sleep a while and retry until some memory is available.	268	or even to sleep a while and retry until some memory is available.
91		269
		270	Example: The following is the C<realloc> function that libev itself uses
		271	which should work with C<realloc> and C<free> functions of all kinds and
		272	is probably a good basis for your own implementation.
		273
		274	static void *
		275	ev_realloc_emul (void *ptr, long size) EV_NOEXCEPT
		276	{
		277	if (size)
		278	return realloc (ptr, size);
		279
		280	free (ptr);
		281	return 0;
		282	}
		283
		284	Example: Replace the libev allocator with one that waits a bit and then
		285	retries.
		286
		287	static void *
		288	persistent_realloc (void *ptr, size_t size)
		289	{
		290	if (!size)
		291	{
		292	free (ptr);
		293	return 0;
		294	}
		295
		296	for (;;)
		297	{
		298	void *newptr = realloc (ptr, size);
		299
		300	if (newptr)
		301	return newptr;
		302
		303	sleep (60);
		304	}
		305	}
		306
		307	...
		308	ev_set_allocator (persistent_realloc);
		309
92	=item ev_set_syserr_cb (void (*cb)(const char *msg));	310	=item ev_set_syserr_cb (void (*cb)(const char *msg) throw ())
93		311
94	Set the callback function to call on a retryable syscall error (such	312	Set the callback function to call on a retryable system call error (such
95	as failed select, poll, epoll_wait). The message is a printable string	313	as failed select, poll, epoll_wait). The message is a printable string
96	indicating the system call or subsystem causing the problem. If this	314	indicating the system call or subsystem causing the problem. If this
97	callback is set, then libev will expect it to remedy the sitution, no	315	callback is set, then libev will expect it to remedy the situation, no
98	matter what, when it returns. That is, libev will generally retry the	316	matter what, when it returns. That is, libev will generally retry the
99	requested operation, or, if the condition doesn't go away, do bad stuff	317	requested operation, or, if the condition doesn't go away, do bad stuff
100	(such as abort).	318	(such as abort).
101		319
		320	Example: This is basically the same thing that libev does internally, too.
		321
		322	static void
		323	fatal_error (const char *msg)
		324	{
		325	perror (msg);
		326	abort ();
		327	}
		328
		329	...
		330	ev_set_syserr_cb (fatal_error);
		331
		332	=item ev_feed_signal (int signum)
		333
		334	This function can be used to "simulate" a signal receive. It is completely
		335	safe to call this function at any time, from any context, including signal
		336	handlers or random threads.
		337
		338	Its main use is to customise signal handling in your process, especially
		339	in the presence of threads. For example, you could block signals
		340	by default in all threads (and specifying C<EVFLAG_NOSIGMASK> when
		341	creating any loops), and in one thread, use C<sigwait> or any other
		342	mechanism to wait for signals, then "deliver" them to libev by calling
		343	C<ev_feed_signal>.
		344
102	=back	345	=back
103		346
104	=head1 FUNCTIONS CONTROLLING THE EVENT LOOP	347	=head1 FUNCTIONS CONTROLLING EVENT LOOPS
105		348
106	An event loop is described by a C<struct ev_loop *>. The library knows two	349	An event loop is described by a C<struct ev_loop *> (the C<struct> is
107	types of such loops, the I<default> loop, which supports signals and child	350	I<not> optional in this case unless libev 3 compatibility is disabled, as
108	events, and dynamically created loops which do not.	351	libev 3 had an C<ev_loop> function colliding with the struct name).
109		352
110	If you use threads, a common model is to run the default event loop	353	The library knows two types of such loops, the I<default> loop, which
111	in your main thread (or in a separate thread) and for each thread you	354	supports child process events, and dynamically created event loops which
112	create, you also create another event loop. Libev itself does no locking	355	do not.
113	whatsoever, so if you mix calls to the same event loop in different
114	threads, make sure you lock (this is usually a bad idea, though, even if
115	done correctly, because it's hideous and inefficient).
116		356
117	=over 4	357	=over 4
118		358
119	=item struct ev_loop *ev_default_loop (unsigned int flags)	359	=item struct ev_loop *ev_default_loop (unsigned int flags)
120		360
121	This will initialise the default event loop if it hasn't been initialised	361	This returns the "default" event loop object, which is what you should
122	yet and return it. If the default loop could not be initialised, returns	362	normally use when you just need "the event loop". Event loop objects and
123	false. If it already was initialised it simply returns it (and ignores the	363	the C<flags> parameter are described in more detail in the entry for
124	flags).	364	C<ev_loop_new>.
		365
		366	If the default loop is already initialised then this function simply
		367	returns it (and ignores the flags. If that is troubling you, check
		368	C<ev_backend ()> afterwards). Otherwise it will create it with the given
		369	flags, which should almost always be C<0>, unless the caller is also the
		370	one calling C<ev_run> or otherwise qualifies as "the main program".
125		371
126	If you don't know what event loop to use, use the one returned from this	372	If you don't know what event loop to use, use the one returned from this
127	function.	373	function (or via the C<EV_DEFAULT> macro).
		374
		375	Note that this function is I<not> thread-safe, so if you want to use it
		376	from multiple threads, you have to employ some kind of mutex (note also
		377	that this case is unlikely, as loops cannot be shared easily between
		378	threads anyway).
		379
		380	The default loop is the only loop that can handle C<ev_child> watchers,
		381	and to do this, it always registers a handler for C<SIGCHLD>. If this is
		382	a problem for your application you can either create a dynamic loop with
		383	C<ev_loop_new> which doesn't do that, or you can simply overwrite the
		384	C<SIGCHLD> signal handler I<after> calling C<ev_default_init>.
		385
		386	Example: This is the most typical usage.
		387
		388	if (!ev_default_loop (0))
		389	fatal ("could not initialise libev, bad $LIBEV_FLAGS in environment?");
		390
		391	Example: Restrict libev to the select and poll backends, and do not allow
		392	environment settings to be taken into account:
		393
		394	ev_default_loop (EVBACKEND_POLL | EVBACKEND_SELECT | EVFLAG_NOENV);
		395
		396	=item struct ev_loop *ev_loop_new (unsigned int flags)
		397
		398	This will create and initialise a new event loop object. If the loop
		399	could not be initialised, returns false.
		400
		401	This function is thread-safe, and one common way to use libev with
		402	threads is indeed to create one loop per thread, and using the default
		403	loop in the "main" or "initial" thread.
128		404
129	The flags argument can be used to specify special behaviour or specific	405	The flags argument can be used to specify special behaviour or specific
130	backends to use, and is usually specified as 0 (or EVFLAG_AUTO).	406	backends to use, and is usually specified as C<0> (or C<EVFLAG_AUTO>).
131		407
132	It supports the following flags:	408	The following flags are supported:
133		409
134	=over 4	410	=over 4
135		411
136	=item C<EVFLAG_AUTO>	412	=item C<EVFLAG_AUTO>
137		413
138	The default flags value. Use this if you have no clue (it's the right	414	The default flags value. Use this if you have no clue (it's the right
139	thing, believe me).	415	thing, believe me).
140		416
141	=item C<EVFLAG_NOENV>	417	=item C<EVFLAG_NOENV>
142		418
143	If this flag bit is ored into the flag value (or the program runs setuid	419	If this flag bit is or'ed into the flag value (or the program runs setuid
144	or setgid) then libev will I<not> look at the environment variable	420	or setgid) then libev will I<not> look at the environment variable
145	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will	421	C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will
146	override the flags completely if it is found in the environment. This is	422	override the flags completely if it is found in the environment. This is
147	useful to try out specific backends to test their performance, or to work	423	useful to try out specific backends to test their performance, to work
148	around bugs.	424	around bugs, or to make libev threadsafe (accessing environment variables
		425	cannot be done in a threadsafe way, but usually it works if no other
		426	thread modifies them).
149		427
		428	=item C<EVFLAG_FORKCHECK>
		429
		430	Instead of calling C<ev_loop_fork> manually after a fork, you can also
		431	make libev check for a fork in each iteration by enabling this flag.
		432
		433	This works by calling C<getpid ()> on every iteration of the loop,
		434	and thus this might slow down your event loop if you do a lot of loop
		435	iterations and little real work, but is usually not noticeable (on my
		436	GNU/Linux system for example, C<getpid> is actually a simple 5-insn
		437	sequence without a system call and thus I<very> fast, but my GNU/Linux
		438	system also has C<pthread_atfork> which is even faster). (Update: glibc
		439	versions 2.25 apparently removed the C<getpid> optimisation again).
		440
		441	The big advantage of this flag is that you can forget about fork (and
		442	forget about forgetting to tell libev about forking, although you still
		443	have to ignore C<SIGPIPE>) when you use this flag.
		444
		445	This flag setting cannot be overridden or specified in the C<LIBEV_FLAGS>
		446	environment variable.
		447
		448	=item C<EVFLAG_NOINOTIFY>
		449
		450	When this flag is specified, then libev will not attempt to use the
		451	I<inotify> API for its C<ev_stat> watchers. Apart from debugging and
		452	testing, this flag can be useful to conserve inotify file descriptors, as
		453	otherwise each loop using C<ev_stat> watchers consumes one inotify handle.
		454
		455	=item C<EVFLAG_SIGNALFD>
		456
		457	When this flag is specified, then libev will attempt to use the
		458	I<signalfd> API for its C<ev_signal> (and C<ev_child>) watchers. This API
		459	delivers signals synchronously, which makes it both faster and might make
		460	it possible to get the queued signal data. It can also simplify signal
		461	handling with threads, as long as you properly block signals in your
		462	threads that are not interested in handling them.
		463
		464	Signalfd will not be used by default as this changes your signal mask, and
		465	there are a lot of shoddy libraries and programs (glib's threadpool for
		466	example) that can't properly initialise their signal masks.
		467
		468	=item C<EVFLAG_NOSIGMASK>
		469
		470	When this flag is specified, then libev will avoid to modify the signal
		471	mask. Specifically, this means you have to make sure signals are unblocked
		472	when you want to receive them.
		473
		474	This behaviour is useful when you want to do your own signal handling, or
		475	want to handle signals only in specific threads and want to avoid libev
		476	unblocking the signals.
		477
		478	It's also required by POSIX in a threaded program, as libev calls
		479	C<sigprocmask>, whose behaviour is officially unspecified.
		480
		481	This flag's behaviour will become the default in future versions of libev.
		482
150	=item C<EVMETHOD_SELECT> (portable select backend)	483	=item C<EVBACKEND_SELECT> (value 1, portable select backend)
151		484
		485	This is your standard select(2) backend. Not I<completely> standard, as
		486	libev tries to roll its own fd_set with no limits on the number of fds,
		487	but if that fails, expect a fairly low limit on the number of fds when
		488	using this backend. It doesn't scale too well (O(highest_fd)), but its
		489	usually the fastest backend for a low number of (low-numbered :) fds.
		490
		491	To get good performance out of this backend you need a high amount of
		492	parallelism (most of the file descriptors should be busy). If you are
		493	writing a server, you should C<accept ()> in a loop to accept as many
		494	connections as possible during one iteration. You might also want to have
		495	a look at C<ev_set_io_collect_interval ()> to increase the amount of
		496	readiness notifications you get per iteration.
		497
		498	This backend maps C<EV_READ> to the C<readfds> set and C<EV_WRITE> to the
		499	C<writefds> set (and to work around Microsoft Windows bugs, also onto the
		500	C<exceptfds> set on that platform).
		501
152	=item C<EVMETHOD_POLL> (poll backend, available everywhere except on windows)	502	=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows)
153		503
154	=item C<EVMETHOD_EPOLL> (linux only)	504	And this is your standard poll(2) backend. It's more complicated
		505	than select, but handles sparse fds better and has no artificial
		506	limit on the number of fds you can use (except it will slow down
		507	considerably with a lot of inactive fds). It scales similarly to select,
		508	i.e. O(total_fds). See the entry for C<EVBACKEND_SELECT>, above, for
		509	performance tips.
155		510
156	=item C<EVMETHOD_KQUEUE> (some bsds only)	511	This backend maps C<EV_READ> to C<POLLIN | POLLERR | POLLHUP>, and
		512	C<EV_WRITE> to C<POLLOUT | POLLERR | POLLHUP>.
157		513
158	=item C<EVMETHOD_DEVPOLL> (solaris 8 only)	514	=item C<EVBACKEND_EPOLL> (value 4, Linux)
159		515
160	=item C<EVMETHOD_PORT> (solaris 10 only)	516	Use the linux-specific epoll(7) interface (for both pre- and post-2.6.9
		517	kernels).
161		518
162	If one or more of these are ored into the flags value, then only these	519	For few fds, this backend is a bit little slower than poll and select, but
163	backends will be tried (in the reverse order as given here). If one are	520	it scales phenomenally better. While poll and select usually scale like
164	specified, any backend will do.	521	O(total_fds) where total_fds is the total number of fds (or the highest
		522	fd), epoll scales either O(1) or O(active_fds).
		523
		524	The epoll mechanism deserves honorable mention as the most misdesigned
		525	of the more advanced event mechanisms: mere annoyances include silently
		526	dropping file descriptors, requiring a system call per change per file
		527	descriptor (and unnecessary guessing of parameters), problems with dup,
		528	returning before the timeout value, resulting in additional iterations
		529	(and only giving 5ms accuracy while select on the same platform gives
		530	0.1ms) and so on. The biggest issue is fork races, however - if a program
		531	forks then I<both> parent and child process have to recreate the epoll
		532	set, which can take considerable time (one syscall per file descriptor)
		533	and is of course hard to detect.
		534
		535	Epoll is also notoriously buggy - embedding epoll fds I<should> work,
		536	but of course I<doesn't>, and epoll just loves to report events for
		537	totally I<different> file descriptors (even already closed ones, so
		538	one cannot even remove them from the set) than registered in the set
		539	(especially on SMP systems). Libev tries to counter these spurious
		540	notifications by employing an additional generation counter and comparing
		541	that against the events to filter out spurious ones, recreating the set
		542	when required. Epoll also erroneously rounds down timeouts, but gives you
		543	no way to know when and by how much, so sometimes you have to busy-wait
		544	because epoll returns immediately despite a nonzero timeout. And last
		545	not least, it also refuses to work with some file descriptors which work
		546	perfectly fine with C<select> (files, many character devices...).
		547
		548	Epoll is truly the train wreck among event poll mechanisms, a frankenpoll,
		549	cobbled together in a hurry, no thought to design or interaction with
		550	others. Oh, the pain, will it ever stop...
		551
		552	While stopping, setting and starting an I/O watcher in the same iteration
		553	will result in some caching, there is still a system call per such
		554	incident (because the same I<file descriptor> could point to a different
		555	I<file description> now), so its best to avoid that. Also, C<dup ()>'ed
		556	file descriptors might not work very well if you register events for both
		557	file descriptors.
		558
		559	Best performance from this backend is achieved by not unregistering all
		560	watchers for a file descriptor until it has been closed, if possible,
		561	i.e. keep at least one watcher active per fd at all times. Stopping and
		562	starting a watcher (without re-setting it) also usually doesn't cause
		563	extra overhead. A fork can both result in spurious notifications as well
		564	as in libev having to destroy and recreate the epoll object, which can
		565	take considerable time and thus should be avoided.
		566
		567	All this means that, in practice, C<EVBACKEND_SELECT> can be as fast or
		568	faster than epoll for maybe up to a hundred file descriptors, depending on
		569	the usage. So sad.
		570
		571	While nominally embeddable in other event loops, this feature is broken in
		572	a lot of kernel revisions, but probably(!) works in current versions.
		573
		574	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		575	C<EVBACKEND_POLL>.
		576
		577	=item C<EVBACKEND_LINUXAIO> (value 64, Linux)
		578
		579	Use the linux-specific linux aio (I<not> C<< aio(7) >> but C<<
		580	io_submit(2) >>) event interface available in post-4.18 kernels (but libev
		581	only tries to use it in 4.19+).
		582
		583	This is another linux trainwreck of an event interface.
		584
		585	If this backend works for you (as of this writing, it was very
		586	experimental), it is the best event interface available on linux and might
		587	be well worth enabling it - if it isn't available in your kernel this will
		588	be detected and this backend will be skipped.
		589
		590	This backend can batch oneshot requests and supports a user-space ring
		591	buffer to receive events. It also doesn't suffer from most of the design
		592	problems of epoll (such as not being able to remove event sources from
		593	the epoll set), and generally sounds too good to be true. Because, this
		594	being the linux kernel, of course it suffers from a whole new set of
		595	limitations, forcing you to fall back to epoll, inheriting all its design
		596	issues.
		597
		598	For one, it is not easily embeddable (but probably could be done using
		599	an event fd at some extra overhead). It also is subject to a system wide
		600	limit that can be configured in F</proc/sys/fs/aio-max-nr>. If no aio
		601	requests are left, this backend will be skipped during initialisation, and
		602	will switch to epoll when the loop is active.
		603
		604	Most problematic in practice, however, is that not all file descriptors
		605	work with it. For example, in linux 5.1, tcp sockets, pipes, event fds,
		606	files, F</dev/null> and a few others are supported, but ttys do not work
		607	properly (a known bug that the kernel developers don't care about, see
		608	L<https://lore.kernel.org/patchwork/patch/1047453/>), so this is not
		609	(yet?) a generic event polling interface.
		610
		611	Overall, it seems the linux developers just don't want it to have a
		612	generic event handling mechanism other than C<select> or C<poll>.
		613
		614	To work around all these problem, the current version of libev uses its
		615	epoll backend as a fallback for file descriptor types that do not work. Or
		616	falls back completely to epoll if the kernel acts up.
		617
		618	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		619	C<EVBACKEND_POLL>.
		620
		621	=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones)
		622
		623	Kqueue deserves special mention, as at the time this backend was
		624	implemented, it was broken on all BSDs except NetBSD (usually it doesn't
		625	work reliably with anything but sockets and pipes, except on Darwin,
		626	where of course it's completely useless). Unlike epoll, however, whose
		627	brokenness is by design, these kqueue bugs can be (and mostly have been)
		628	fixed without API changes to existing programs. For this reason it's not
		629	being "auto-detected" on all platforms unless you explicitly specify it
		630	in the flags (i.e. using C<EVBACKEND_KQUEUE>) or libev was compiled on a
		631	known-to-be-good (-enough) system like NetBSD.
		632
		633	You still can embed kqueue into a normal poll or select backend and use it
		634	only for sockets (after having made sure that sockets work with kqueue on
		635	the target platform). See C<ev_embed> watchers for more info.
		636
		637	It scales in the same way as the epoll backend, but the interface to the
		638	kernel is more efficient (which says nothing about its actual speed, of
		639	course). While stopping, setting and starting an I/O watcher does never
		640	cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to
		641	two event changes per incident. Support for C<fork ()> is very bad (you
		642	might have to leak fd's on fork, but it's more sane than epoll) and it
		643	drops fds silently in similarly hard-to-detect cases.
		644
		645	This backend usually performs well under most conditions.
		646
		647	While nominally embeddable in other event loops, this doesn't work
		648	everywhere, so you might need to test for this. And since it is broken
		649	almost everywhere, you should only use it when you have a lot of sockets
		650	(for which it usually works), by embedding it into another event loop
		651	(e.g. C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> (but C<poll> is of course
		652	also broken on OS X)) and, did I mention it, using it only for sockets.
		653
		654	This backend maps C<EV_READ> into an C<EVFILT_READ> kevent with
		655	C<NOTE_EOF>, and C<EV_WRITE> into an C<EVFILT_WRITE> kevent with
		656	C<NOTE_EOF>.
		657
		658	=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8)
		659
		660	This is not implemented yet (and might never be, unless you send me an
		661	implementation). According to reports, C</dev/poll> only supports sockets
		662	and is not embeddable, which would limit the usefulness of this backend
		663	immensely.
		664
		665	=item C<EVBACKEND_PORT> (value 32, Solaris 10)
		666
		667	This uses the Solaris 10 event port mechanism. As with everything on Solaris,
		668	it's really slow, but it still scales very well (O(active_fds)).
		669
		670	While this backend scales well, it requires one system call per active
		671	file descriptor per loop iteration. For small and medium numbers of file
		672	descriptors a "slow" C<EVBACKEND_SELECT> or C<EVBACKEND_POLL> backend
		673	might perform better.
		674
		675	On the positive side, this backend actually performed fully to
		676	specification in all tests and is fully embeddable, which is a rare feat
		677	among the OS-specific backends (I vastly prefer correctness over speed
		678	hacks).
		679
		680	On the negative side, the interface is I<bizarre> - so bizarre that
		681	even sun itself gets it wrong in their code examples: The event polling
		682	function sometimes returns events to the caller even though an error
		683	occurred, but with no indication whether it has done so or not (yes, it's
		684	even documented that way) - deadly for edge-triggered interfaces where you
		685	absolutely have to know whether an event occurred or not because you have
		686	to re-arm the watcher.
		687
		688	Fortunately libev seems to be able to work around these idiocies.
		689
		690	This backend maps C<EV_READ> and C<EV_WRITE> in the same way as
		691	C<EVBACKEND_POLL>.
		692
		693	=item C<EVBACKEND_ALL>
		694
		695	Try all backends (even potentially broken ones that wouldn't be tried
		696	with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as
		697	C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>.
		698
		699	It is definitely not recommended to use this flag, use whatever
		700	C<ev_recommended_backends ()> returns, or simply do not specify a backend
		701	at all.
		702
		703	=item C<EVBACKEND_MASK>
		704
		705	Not a backend at all, but a mask to select all backend bits from a
		706	C<flags> value, in case you want to mask out any backends from a flags
		707	value (e.g. when modifying the C<LIBEV_FLAGS> environment variable).
165		708
166	=back	709	=back
167		710
168	=item struct ev_loop *ev_loop_new (unsigned int flags)	711	If one or more of the backend flags are or'ed into the flags value,
		712	then only these backends will be tried (in the reverse order as listed
		713	here). If none are specified, all backends in C<ev_recommended_backends
		714	()> will be tried.
169		715
170	Similar to C<ev_default_loop>, but always creates a new event loop that is	716	Example: Try to create a event loop that uses epoll and nothing else.
171	always distinct from the default loop. Unlike the default loop, it cannot
172	handle signal and child watchers, and attempts to do so will be greeted by
173	undefined behaviour (or a failed assertion if assertions are enabled).
174		717
175	=item ev_default_destroy ()	718	struct ev_loop *epoller = ev_loop_new (EVBACKEND_EPOLL | EVFLAG_NOENV);
		719	if (!epoller)
		720	fatal ("no epoll found here, maybe it hides under your chair");
176		721
177	Destroys the default loop again (frees all memory and kernel state	722	Example: Use whatever libev has to offer, but make sure that kqueue is
178	etc.). This stops all registered event watchers (by not touching them in	723	used if available.
179	any way whatsoever, although you cannot rely on this :).	724
		725	struct ev_loop *loop = ev_loop_new (ev_recommended_backends () | EVBACKEND_KQUEUE);
		726
		727	Example: Similarly, on linux, you mgiht want to take advantage of the
		728	linux aio backend if possible, but fall back to something else if that
		729	isn't available.
		730
		731	struct ev_loop *loop = ev_loop_new (ev_recommended_backends () | EVBACKEND_LINUXAIO);
180		732
181	=item ev_loop_destroy (loop)	733	=item ev_loop_destroy (loop)
182		734
183	Like C<ev_default_destroy>, but destroys an event loop created by an	735	Destroys an event loop object (frees all memory and kernel state
184	earlier call to C<ev_loop_new>.	736	etc.). None of the active event watchers will be stopped in the normal
		737	sense, so e.g. C<ev_is_active> might still return true. It is your
		738	responsibility to either stop all watchers cleanly yourself I<before>
		739	calling this function, or cope with the fact afterwards (which is usually
		740	the easiest thing, you can just ignore the watchers and/or C<free ()> them
		741	for example).
185		742
186	=item ev_default_fork ()	743	Note that certain global state, such as signal state (and installed signal
		744	handlers), will not be freed by this function, and related watchers (such
		745	as signal and child watchers) would need to be stopped manually.
187		746
		747	This function is normally used on loop objects allocated by
		748	C<ev_loop_new>, but it can also be used on the default loop returned by
		749	C<ev_default_loop>, in which case it is not thread-safe.
		750
		751	Note that it is not advisable to call this function on the default loop
		752	except in the rare occasion where you really need to free its resources.
		753	If you need dynamically allocated loops it is better to use C<ev_loop_new>
		754	and C<ev_loop_destroy>.
		755
		756	=item ev_loop_fork (loop)
		757
		758	This function sets a flag that causes subsequent C<ev_run> iterations
188	This function reinitialises the kernel state for backends that have	759	to reinitialise the kernel state for backends that have one. Despite
189	one. Despite the name, you can call it anytime, but it makes most sense	760	the name, you can call it anytime you are allowed to start or stop
190	after forking, in either the parent or child process (or both, but that	761	watchers (except inside an C<ev_prepare> callback), but it makes most
191	again makes little sense).	762	sense after forking, in the child process. You I<must> call it (or use
		763	C<EVFLAG_FORKCHECK>) in the child before resuming or calling C<ev_run>.
192		764
193	You I<must> call this function after forking if and only if you want to	765	In addition, if you want to reuse a loop (via this function or
194	use the event library in both processes. If you just fork+exec, you don't	766	C<EVFLAG_FORKCHECK>), you I<also> have to ignore C<SIGPIPE>.
195	have to call it.	767
		768	Again, you I<have> to call it on I<any> loop that you want to re-use after
		769	a fork, I<even if you do not plan to use the loop in the parent>. This is
		770	because some kernel interfaces *cough* I<kqueue> *cough* do funny things
		771	during fork.
		772
		773	On the other hand, you only need to call this function in the child
		774	process if and only if you want to use the event loop in the child. If
		775	you just fork+exec or create a new loop in the child, you don't have to
		776	call it at all (in fact, C<epoll> is so badly broken that it makes a
		777	difference, but libev will usually detect this case on its own and do a
		778	costly reset of the backend).
196		779
197	The function itself is quite fast and it's usually not a problem to call	780	The function itself is quite fast and it's usually not a problem to call
198	it just in case after a fork. To make this easy, the function will fit in	781	it just in case after a fork.
199	quite nicely into a call to C<pthread_atfork>:
200		782
		783	Example: Automate calling C<ev_loop_fork> on the default loop when
		784	using pthreads.
		785
		786	static void
		787	post_fork_child (void)
		788	{
		789	ev_loop_fork (EV_DEFAULT);
		790	}
		791
		792	...
201	pthread_atfork (0, 0, ev_default_fork);	793	pthread_atfork (0, 0, post_fork_child);
202		794
203	=item ev_loop_fork (loop)	795	=item int ev_is_default_loop (loop)
204		796
205	Like C<ev_default_fork>, but acts on an event loop created by	797	Returns true when the given loop is, in fact, the default loop, and false
206	C<ev_loop_new>. Yes, you have to call this on every allocated event loop	798	otherwise.
207	after fork, and how you do this is entirely your own problem.
208		799
		800	=item unsigned int ev_iteration (loop)
		801
		802	Returns the current iteration count for the event loop, which is identical
		803	to the number of times libev did poll for new events. It starts at C<0>
		804	and happily wraps around with enough iterations.
		805
		806	This value can sometimes be useful as a generation counter of sorts (it
		807	"ticks" the number of loop iterations), as it roughly corresponds with
		808	C<ev_prepare> and C<ev_check> calls - and is incremented between the
		809	prepare and check phases.
		810
209	=item unsigned int ev_method (loop)	811	=item unsigned int ev_depth (loop)
210		812
		813	Returns the number of times C<ev_run> was entered minus the number of
		814	times C<ev_run> was exited normally, in other words, the recursion depth.
		815
		816	Outside C<ev_run>, this number is zero. In a callback, this number is
		817	C<1>, unless C<ev_run> was invoked recursively (or from another thread),
		818	in which case it is higher.
		819
		820	Leaving C<ev_run> abnormally (setjmp/longjmp, cancelling the thread,
		821	throwing an exception etc.), doesn't count as "exit" - consider this
		822	as a hint to avoid such ungentleman-like behaviour unless it's really
		823	convenient, in which case it is fully supported.
		824
		825	=item unsigned int ev_backend (loop)
		826
211	Returns one of the C<EVMETHOD_*> flags indicating the event backend in	827	Returns one of the C<EVBACKEND_*> flags indicating the event backend in
212	use.	828	use.
213		829
214	=item ev_tstamp ev_now (loop)	830	=item ev_tstamp ev_now (loop)
215		831
216	Returns the current "event loop time", which is the time the event loop	832	Returns the current "event loop time", which is the time the event loop
217	got events and started processing them. This timestamp does not change	833	received events and started processing them. This timestamp does not
218	as long as callbacks are being processed, and this is also the base time	834	change as long as callbacks are being processed, and this is also the base
219	used for relative timers. You can treat it as the timestamp of the event	835	time used for relative timers. You can treat it as the timestamp of the
220	occuring (or more correctly, the mainloop finding out about it).	836	event occurring (or more correctly, libev finding out about it).
221		837
		838	=item ev_now_update (loop)
		839
		840	Establishes the current time by querying the kernel, updating the time
		841	returned by C<ev_now ()> in the progress. This is a costly operation and
		842	is usually done automatically within C<ev_run ()>.
		843
		844	This function is rarely useful, but when some event callback runs for a
		845	very long time without entering the event loop, updating libev's idea of
		846	the current time is a good idea.
		847
		848	See also L</The special problem of time updates> in the C<ev_timer> section.
		849
		850	=item ev_suspend (loop)
		851
		852	=item ev_resume (loop)
		853
		854	These two functions suspend and resume an event loop, for use when the
		855	loop is not used for a while and timeouts should not be processed.
		856
		857	A typical use case would be an interactive program such as a game: When
		858	the user presses C<^Z> to suspend the game and resumes it an hour later it
		859	would be best to handle timeouts as if no time had actually passed while
		860	the program was suspended. This can be achieved by calling C<ev_suspend>
		861	in your C<SIGTSTP> handler, sending yourself a C<SIGSTOP> and calling
		862	C<ev_resume> directly afterwards to resume timer processing.
		863
		864	Effectively, all C<ev_timer> watchers will be delayed by the time spend
		865	between C<ev_suspend> and C<ev_resume>, and all C<ev_periodic> watchers
		866	will be rescheduled (that is, they will lose any events that would have
		867	occurred while suspended).
		868
		869	After calling C<ev_suspend> you B<must not> call I<any> function on the
		870	given loop other than C<ev_resume>, and you B<must not> call C<ev_resume>
		871	without a previous call to C<ev_suspend>.
		872
		873	Calling C<ev_suspend>/C<ev_resume> has the side effect of updating the
		874	event loop time (see C<ev_now_update>).
		875
222	=item ev_loop (loop, int flags)	876	=item bool ev_run (loop, int flags)
223		877
224	Finally, this is it, the event handler. This function usually is called	878	Finally, this is it, the event handler. This function usually is called
225	after you initialised all your watchers and you want to start handling	879	after you have initialised all your watchers and you want to start
226	events.	880	handling events. It will ask the operating system for any new events, call
		881	the watcher callbacks, and then repeat the whole process indefinitely: This
		882	is why event loops are called I<loops>.
227		883
228	If the flags argument is specified as 0, it will not return until either	884	If the flags argument is specified as C<0>, it will keep handling events
229	no event watchers are active anymore or C<ev_unloop> was called.	885	until either no event watchers are active anymore or C<ev_break> was
		886	called.
230		887
		888	The return value is false if there are no more active watchers (which
		889	usually means "all jobs done" or "deadlock"), and true in all other cases
		890	(which usually means " you should call C<ev_run> again").
		891
		892	Please note that an explicit C<ev_break> is usually better than
		893	relying on all watchers to be stopped when deciding when a program has
		894	finished (especially in interactive programs), but having a program
		895	that automatically loops as long as it has to and no longer by virtue
		896	of relying on its watchers stopping correctly, that is truly a thing of
		897	beauty.
		898
		899	This function is I<mostly> exception-safe - you can break out of a
		900	C<ev_run> call by calling C<longjmp> in a callback, throwing a C++
		901	exception and so on. This does not decrement the C<ev_depth> value, nor
		902	will it clear any outstanding C<EVBREAK_ONE> breaks.
		903
231	A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle	904	A flags value of C<EVRUN_NOWAIT> will look for new events, will handle
232	those events and any outstanding ones, but will not block your process in	905	those events and any already outstanding ones, but will not wait and
233	case there are no events and will return after one iteration of the loop.	906	block your process in case there are no events and will return after one
		907	iteration of the loop. This is sometimes useful to poll and handle new
		908	events while doing lengthy calculations, to keep the program responsive.
234		909
235	A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if	910	A flags value of C<EVRUN_ONCE> will look for new events (waiting if
236	neccessary) and will handle those and any outstanding ones. It will block	911	necessary) and will handle those and any already outstanding ones. It
237	your process until at least one new event arrives, and will return after	912	will block your process until at least one new event arrives (which could
		913	be an event internal to libev itself, so there is no guarantee that a
		914	user-registered callback will be called), and will return after one
238	one iteration of the loop.	915	iteration of the loop.
239		916
240	This flags value could be used to implement alternative looping	917	This is useful if you are waiting for some external event in conjunction
241	constructs, but the C<prepare> and C<check> watchers provide a better and	918	with something not expressible using other libev watchers (i.e. "roll your
242	more generic mechanism.	919	own C<ev_run>"). However, a pair of C<ev_prepare>/C<ev_check> watchers is
		920	usually a better approach for this kind of thing.
243		921
244	Here are the gory details of what ev_loop does:	922	Here are the gory details of what C<ev_run> does (this is for your
		923	understanding, not a guarantee that things will work exactly like this in
		924	future versions):
245		925
246	1. If there are no active watchers (reference count is zero), return.	926	- Increment loop depth.
		927	- Reset the ev_break status.
		928	- Before the first iteration, call any pending watchers.
		929	LOOP:
		930	- If EVFLAG_FORKCHECK was used, check for a fork.
		931	- If a fork was detected (by any means), queue and call all fork watchers.
247	2. Queue and immediately call all prepare watchers.	932	- Queue and call all prepare watchers.
		933	- If ev_break was called, goto FINISH.
248	3. If we have been forked, recreate the kernel state.	934	- If we have been forked, detach and recreate the kernel state
		935	as to not disturb the other process.
249	4. Update the kernel state with all outstanding changes.	936	- Update the kernel state with all outstanding changes.
250	5. Update the "event loop time".	937	- Update the "event loop time" (ev_now ()).
251	6. Calculate for how long to block.	938	- Calculate for how long to sleep or block, if at all
		939	(active idle watchers, EVRUN_NOWAIT or not having
		940	any active watchers at all will result in not sleeping).
		941	- Sleep if the I/O and timer collect interval say so.
		942	- Increment loop iteration counter.
252	7. Block the process, waiting for events.	943	- Block the process, waiting for any events.
		944	- Queue all outstanding I/O (fd) events.
253	8. Update the "event loop time" and do time jump handling.	945	- Update the "event loop time" (ev_now ()), and do time jump adjustments.
254	9. Queue all outstanding timers.	946	- Queue all expired timers.
255	10. Queue all outstanding periodics.	947	- Queue all expired periodics.
256	11. If no events are pending now, queue all idle watchers.	948	- Queue all idle watchers with priority higher than that of pending events.
257	12. Queue all check watchers.	949	- Queue all check watchers.
258	13. Call all queued watchers in reverse order (i.e. check watchers first).	950	- Call all queued watchers in reverse order (i.e. check watchers first).
259	14. If ev_unloop has been called or EVLOOP_ONESHOT or EVLOOP_NONBLOCK	951	Signals and child watchers are implemented as I/O watchers, and will
260	was used, return, otherwise continue with step #1.	952	be handled here by queueing them when their watcher gets executed.
		953	- If ev_break has been called, or EVRUN_ONCE or EVRUN_NOWAIT
		954	were used, or there are no active watchers, goto FINISH, otherwise
		955	continue with step LOOP.
		956	FINISH:
		957	- Reset the ev_break status iff it was EVBREAK_ONE.
		958	- Decrement the loop depth.
		959	- Return.
261		960
		961	Example: Queue some jobs and then loop until no events are outstanding
		962	anymore.
		963
		964	... queue jobs here, make sure they register event watchers as long
		965	... as they still have work to do (even an idle watcher will do..)
		966	ev_run (my_loop, 0);
		967	... jobs done or somebody called break. yeah!
		968
262	=item ev_unloop (loop, how)	969	=item ev_break (loop, how)
263		970
264	Can be used to make a call to C<ev_loop> return early (but only after it	971	Can be used to make a call to C<ev_run> return early (but only after it
265	has processed all outstanding events). The C<how> argument must be either	972	has processed all outstanding events). The C<how> argument must be either
266	C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or	973	C<EVBREAK_ONE>, which will make the innermost C<ev_run> call return, or
267	C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return.	974	C<EVBREAK_ALL>, which will make all nested C<ev_run> calls return.
		975
		976	This "break state" will be cleared on the next call to C<ev_run>.
		977
		978	It is safe to call C<ev_break> from outside any C<ev_run> calls, too, in
		979	which case it will have no effect.
268		980
269	=item ev_ref (loop)	981	=item ev_ref (loop)
270		982
271	=item ev_unref (loop)	983	=item ev_unref (loop)
272		984
273	Ref/unref can be used to add or remove a reference count on the event	985	Ref/unref can be used to add or remove a reference count on the event
274	loop: Every watcher keeps one reference, and as long as the reference	986	loop: Every watcher keeps one reference, and as long as the reference
275	count is nonzero, C<ev_loop> will not return on its own. If you have	987	count is nonzero, C<ev_run> will not return on its own.
276	a watcher you never unregister that should not keep C<ev_loop> from	988
277	returning, ev_unref() after starting, and ev_ref() before stopping it. For	989	This is useful when you have a watcher that you never intend to
		990	unregister, but that nevertheless should not keep C<ev_run> from
		991	returning. In such a case, call C<ev_unref> after starting, and C<ev_ref>
		992	before stopping it.
		993
278	example, libev itself uses this for its internal signal pipe: It is not	994	As an example, libev itself uses this for its internal signal pipe: It
279	visible to the libev user and should not keep C<ev_loop> from exiting if	995	is not visible to the libev user and should not keep C<ev_run> from
280	no event watchers registered by it are active. It is also an excellent	996	exiting if no event watchers registered by it are active. It is also an
281	way to do this for generic recurring timers or from within third-party	997	excellent way to do this for generic recurring timers or from within
282	libraries. Just remember to I<unref after start> and I<ref before stop>.	998	third-party libraries. Just remember to I<unref after start> and I<ref
		999	before stop> (but only if the watcher wasn't active before, or was active
		1000	before, respectively. Note also that libev might stop watchers itself
		1001	(e.g. non-repeating timers) in which case you have to C<ev_ref>
		1002	in the callback).
		1003
		1004	Example: Create a signal watcher, but keep it from keeping C<ev_run>
		1005	running when nothing else is active.
		1006
		1007	ev_signal exitsig;
		1008	ev_signal_init (&exitsig, sig_cb, SIGINT);
		1009	ev_signal_start (loop, &exitsig);
		1010	ev_unref (loop);
		1011
		1012	Example: For some weird reason, unregister the above signal handler again.
		1013
		1014	ev_ref (loop);
		1015	ev_signal_stop (loop, &exitsig);
		1016
		1017	=item ev_set_io_collect_interval (loop, ev_tstamp interval)
		1018
		1019	=item ev_set_timeout_collect_interval (loop, ev_tstamp interval)
		1020
		1021	These advanced functions influence the time that libev will spend waiting
		1022	for events. Both time intervals are by default C<0>, meaning that libev
		1023	will try to invoke timer/periodic callbacks and I/O callbacks with minimum
		1024	latency.
		1025
		1026	Setting these to a higher value (the C<interval> I<must> be >= C<0>)
		1027	allows libev to delay invocation of I/O and timer/periodic callbacks
		1028	to increase efficiency of loop iterations (or to increase power-saving
		1029	opportunities).
		1030
		1031	The idea is that sometimes your program runs just fast enough to handle
		1032	one (or very few) event(s) per loop iteration. While this makes the
		1033	program responsive, it also wastes a lot of CPU time to poll for new
		1034	events, especially with backends like C<select ()> which have a high
		1035	overhead for the actual polling but can deliver many events at once.
		1036
		1037	By setting a higher I<io collect interval> you allow libev to spend more
		1038	time collecting I/O events, so you can handle more events per iteration,
		1039	at the cost of increasing latency. Timeouts (both C<ev_periodic> and
		1040	C<ev_timer>) will not be affected. Setting this to a non-null value will
		1041	introduce an additional C<ev_sleep ()> call into most loop iterations. The
		1042	sleep time ensures that libev will not poll for I/O events more often then
		1043	once per this interval, on average (as long as the host time resolution is
		1044	good enough).
		1045
		1046	Likewise, by setting a higher I<timeout collect interval> you allow libev
		1047	to spend more time collecting timeouts, at the expense of increased
		1048	latency/jitter/inexactness (the watcher callback will be called
		1049	later). C<ev_io> watchers will not be affected. Setting this to a non-null
		1050	value will not introduce any overhead in libev.
		1051
		1052	Many (busy) programs can usually benefit by setting the I/O collect
		1053	interval to a value near C<0.1> or so, which is often enough for
		1054	interactive servers (of course not for games), likewise for timeouts. It
		1055	usually doesn't make much sense to set it to a lower value than C<0.01>,
		1056	as this approaches the timing granularity of most systems. Note that if
		1057	you do transactions with the outside world and you can't increase the
		1058	parallelity, then this setting will limit your transaction rate (if you
		1059	need to poll once per transaction and the I/O collect interval is 0.01,
		1060	then you can't do more than 100 transactions per second).
		1061
		1062	Setting the I<timeout collect interval> can improve the opportunity for
		1063	saving power, as the program will "bundle" timer callback invocations that
		1064	are "near" in time together, by delaying some, thus reducing the number of
		1065	times the process sleeps and wakes up again. Another useful technique to
		1066	reduce iterations/wake-ups is to use C<ev_periodic> watchers and make sure
		1067	they fire on, say, one-second boundaries only.
		1068
		1069	Example: we only need 0.1s timeout granularity, and we wish not to poll
		1070	more often than 100 times per second:
		1071
		1072	ev_set_timeout_collect_interval (EV_DEFAULT_UC_ 0.1);
		1073	ev_set_io_collect_interval (EV_DEFAULT_UC_ 0.01);
		1074
		1075	=item ev_invoke_pending (loop)
		1076
		1077	This call will simply invoke all pending watchers while resetting their
		1078	pending state. Normally, C<ev_run> does this automatically when required,
		1079	but when overriding the invoke callback this call comes handy. This
		1080	function can be invoked from a watcher - this can be useful for example
		1081	when you want to do some lengthy calculation and want to pass further
		1082	event handling to another thread (you still have to make sure only one
		1083	thread executes within C<ev_invoke_pending> or C<ev_run> of course).
		1084
		1085	=item int ev_pending_count (loop)
		1086
		1087	Returns the number of pending watchers - zero indicates that no watchers
		1088	are pending.
		1089
		1090	=item ev_set_invoke_pending_cb (loop, void (*invoke_pending_cb)(EV_P))
		1091
		1092	This overrides the invoke pending functionality of the loop: Instead of
		1093	invoking all pending watchers when there are any, C<ev_run> will call
		1094	this callback instead. This is useful, for example, when you want to
		1095	invoke the actual watchers inside another context (another thread etc.).
		1096
		1097	If you want to reset the callback, use C<ev_invoke_pending> as new
		1098	callback.
		1099
		1100	=item ev_set_loop_release_cb (loop, void (*release)(EV_P) throw (), void (*acquire)(EV_P) throw ())
		1101
		1102	Sometimes you want to share the same loop between multiple threads. This
		1103	can be done relatively simply by putting mutex_lock/unlock calls around
		1104	each call to a libev function.
		1105
		1106	However, C<ev_run> can run an indefinite time, so it is not feasible
		1107	to wait for it to return. One way around this is to wake up the event
		1108	loop via C<ev_break> and C<ev_async_send>, another way is to set these
		1109	I<release> and I<acquire> callbacks on the loop.
		1110
		1111	When set, then C<release> will be called just before the thread is
		1112	suspended waiting for new events, and C<acquire> is called just
		1113	afterwards.
		1114
		1115	Ideally, C<release> will just call your mutex_unlock function, and
		1116	C<acquire> will just call the mutex_lock function again.
		1117
		1118	While event loop modifications are allowed between invocations of
		1119	C<release> and C<acquire> (that's their only purpose after all), no
		1120	modifications done will affect the event loop, i.e. adding watchers will
		1121	have no effect on the set of file descriptors being watched, or the time
		1122	waited. Use an C<ev_async> watcher to wake up C<ev_run> when you want it
		1123	to take note of any changes you made.
		1124
		1125	In theory, threads executing C<ev_run> will be async-cancel safe between
		1126	invocations of C<release> and C<acquire>.
		1127
		1128	See also the locking example in the C<THREADS> section later in this
		1129	document.
		1130
		1131	=item ev_set_userdata (loop, void *data)
		1132
		1133	=item void *ev_userdata (loop)
		1134
		1135	Set and retrieve a single C<void *> associated with a loop. When
		1136	C<ev_set_userdata> has never been called, then C<ev_userdata> returns
		1137	C<0>.
		1138
		1139	These two functions can be used to associate arbitrary data with a loop,
		1140	and are intended solely for the C<invoke_pending_cb>, C<release> and
		1141	C<acquire> callbacks described above, but of course can be (ab-)used for
		1142	any other purpose as well.
		1143
		1144	=item ev_verify (loop)
		1145
		1146	This function only does something when C<EV_VERIFY> support has been
		1147	compiled in, which is the default for non-minimal builds. It tries to go
		1148	through all internal structures and checks them for validity. If anything
		1149	is found to be inconsistent, it will print an error message to standard
		1150	error and call C<abort ()>.
		1151
		1152	This can be used to catch bugs inside libev itself: under normal
		1153	circumstances, this function will never abort as of course libev keeps its
		1154	data structures consistent.
283		1155
284	=back	1156	=back
285		1157
		1158
286	=head1 ANATOMY OF A WATCHER	1159	=head1 ANATOMY OF A WATCHER
287		1160
		1161	In the following description, uppercase C<TYPE> in names stands for the
		1162	watcher type, e.g. C<ev_TYPE_start> can mean C<ev_timer_start> for timer
		1163	watchers and C<ev_io_start> for I/O watchers.
		1164
288	A watcher is a structure that you create and register to record your	1165	A watcher is an opaque structure that you allocate and register to record
289	interest in some event. For instance, if you want to wait for STDIN to	1166	your interest in some event. To make a concrete example, imagine you want
290	become readable, you would create an C<ev_io> watcher for that:	1167	to wait for STDIN to become readable, you would create an C<ev_io> watcher
		1168	for that:
291		1169
292	static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)	1170	static void my_cb (struct ev_loop *loop, ev_io *w, int revents)
293	{	1171	{
294	ev_io_stop (w);	1172	ev_io_stop (w);
295	ev_unloop (loop, EVUNLOOP_ALL);	1173	ev_break (loop, EVBREAK_ALL);
296	}	1174	}
297		1175
298	struct ev_loop *loop = ev_default_loop (0);	1176	struct ev_loop *loop = ev_default_loop (0);
		1177
299	struct ev_io stdin_watcher;	1178	ev_io stdin_watcher;
		1179
300	ev_init (&stdin_watcher, my_cb);	1180	ev_init (&stdin_watcher, my_cb);
301	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);	1181	ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
302	ev_io_start (loop, &stdin_watcher);	1182	ev_io_start (loop, &stdin_watcher);
		1183
303	ev_loop (loop, 0);	1184	ev_run (loop, 0);
304		1185
305	As you can see, you are responsible for allocating the memory for your	1186	As you can see, you are responsible for allocating the memory for your
306	watcher structures (and it is usually a bad idea to do this on the stack,	1187	watcher structures (and it is I<usually> a bad idea to do this on the
307	although this can sometimes be quite valid).	1188	stack).
308		1189
		1190	Each watcher has an associated watcher structure (called C<struct ev_TYPE>
		1191	or simply C<ev_TYPE>, as typedefs are provided for all watcher structs).
		1192
309	Each watcher structure must be initialised by a call to C<ev_init	1193	Each watcher structure must be initialised by a call to C<ev_init (watcher
310	(watcher *, callback)>, which expects a callback to be provided. This	1194	*, callback)>, which expects a callback to be provided. This callback is
311	callback gets invoked each time the event occurs (or, in the case of io	1195	invoked each time the event occurs (or, in the case of I/O watchers, each
312	watchers, each time the event loop detects that the file descriptor given	1196	time the event loop detects that the file descriptor given is readable
313	is readable and/or writable).	1197	and/or writable).
314		1198
315	Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro	1199	Each watcher type further has its own C<< ev_TYPE_set (watcher *, ...) >>
316	with arguments specific to this watcher type. There is also a macro	1200	macro to configure it, with arguments specific to the watcher type. There
317	to combine initialisation and setting in one call: C<< ev_<type>_init	1201	is also a macro to combine initialisation and setting in one call: C<<
318	(watcher *, callback, ...) >>.	1202	ev_TYPE_init (watcher *, callback, ...) >>.
319		1203
320	To make the watcher actually watch out for events, you have to start it	1204	To make the watcher actually watch out for events, you have to start it
321	with a watcher-specific start function (C<< ev_<type>_start (loop, watcher	1205	with a watcher-specific start function (C<< ev_TYPE_start (loop, watcher
322	*) >>), and you can stop watching for events at any time by calling the	1206	*) >>), and you can stop watching for events at any time by calling the
323	corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>.	1207	corresponding stop function (C<< ev_TYPE_stop (loop, watcher *) >>.
324		1208
325	As long as your watcher is active (has been started but not stopped) you	1209	As long as your watcher is active (has been started but not stopped) you
326	must not touch the values stored in it. Most specifically you must never	1210	must not touch the values stored in it. Most specifically you must never
327	reinitialise it or call its set method.	1211	reinitialise it or call its C<ev_TYPE_set> macro.
328
329	You can check whether an event is active by calling the C<ev_is_active
330	(watcher *)> macro. To see whether an event is outstanding (but the
331	callback for it has not been called yet) you can use the C<ev_is_pending
332	(watcher *)> macro.
333		1212
334	Each and every callback receives the event loop pointer as first, the	1213	Each and every callback receives the event loop pointer as first, the
335	registered watcher structure as second, and a bitset of received events as	1214	registered watcher structure as second, and a bitset of received events as
336	third argument.	1215	third argument.
337		1216
…		…
346	=item C<EV_WRITE>	1225	=item C<EV_WRITE>
347		1226
348	The file descriptor in the C<ev_io> watcher has become readable and/or	1227	The file descriptor in the C<ev_io> watcher has become readable and/or
349	writable.	1228	writable.
350		1229
351	=item C<EV_TIMEOUT>	1230	=item C<EV_TIMER>
352		1231
353	The C<ev_timer> watcher has timed out.	1232	The C<ev_timer> watcher has timed out.
354		1233
355	=item C<EV_PERIODIC>	1234	=item C<EV_PERIODIC>
356		1235
…		…
362		1241
363	=item C<EV_CHILD>	1242	=item C<EV_CHILD>
364		1243
365	The pid specified in the C<ev_child> watcher has received a status change.	1244	The pid specified in the C<ev_child> watcher has received a status change.
366		1245
		1246	=item C<EV_STAT>
		1247
		1248	The path specified in the C<ev_stat> watcher changed its attributes somehow.
		1249
367	=item C<EV_IDLE>	1250	=item C<EV_IDLE>
368		1251
369	The C<ev_idle> watcher has determined that you have nothing better to do.	1252	The C<ev_idle> watcher has determined that you have nothing better to do.
370		1253
371	=item C<EV_PREPARE>	1254	=item C<EV_PREPARE>
372		1255
373	=item C<EV_CHECK>	1256	=item C<EV_CHECK>
374		1257
375	All C<ev_prepare> watchers are invoked just I<before> C<ev_loop> starts	1258	All C<ev_prepare> watchers are invoked just I<before> C<ev_run> starts to
376	to gather new events, and all C<ev_check> watchers are invoked just after	1259	gather new events, and all C<ev_check> watchers are queued (not invoked)
377	C<ev_loop> has gathered them, but before it invokes any callbacks for any	1260	just after C<ev_run> has gathered them, but before it queues any callbacks
		1261	for any received events. That means C<ev_prepare> watchers are the last
		1262	watchers invoked before the event loop sleeps or polls for new events, and
		1263	C<ev_check> watchers will be invoked before any other watchers of the same
		1264	or lower priority within an event loop iteration.
		1265
378	received events. Callbacks of both watcher types can start and stop as	1266	Callbacks of both watcher types can start and stop as many watchers as
379	many watchers as they want, and all of them will be taken into account	1267	they want, and all of them will be taken into account (for example, a
380	(for example, a C<ev_prepare> watcher might start an idle watcher to keep	1268	C<ev_prepare> watcher might start an idle watcher to keep C<ev_run> from
381	C<ev_loop> from blocking).	1269	blocking).
		1270
		1271	=item C<EV_EMBED>
		1272
		1273	The embedded event loop specified in the C<ev_embed> watcher needs attention.
		1274
		1275	=item C<EV_FORK>
		1276
		1277	The event loop has been resumed in the child process after fork (see
		1278	C<ev_fork>).
		1279
		1280	=item C<EV_CLEANUP>
		1281
		1282	The event loop is about to be destroyed (see C<ev_cleanup>).
		1283
		1284	=item C<EV_ASYNC>
		1285
		1286	The given async watcher has been asynchronously notified (see C<ev_async>).
		1287
		1288	=item C<EV_CUSTOM>
		1289
		1290	Not ever sent (or otherwise used) by libev itself, but can be freely used
		1291	by libev users to signal watchers (e.g. via C<ev_feed_event>).
382		1292
383	=item C<EV_ERROR>	1293	=item C<EV_ERROR>
384		1294
385	An unspecified error has occured, the watcher has been stopped. This might	1295	An unspecified error has occurred, the watcher has been stopped. This might
386	happen because the watcher could not be properly started because libev	1296	happen because the watcher could not be properly started because libev
387	ran out of memory, a file descriptor was found to be closed or any other	1297	ran out of memory, a file descriptor was found to be closed or any other
		1298	problem. Libev considers these application bugs.
		1299
388	problem. You best act on it by reporting the problem and somehow coping	1300	You best act on it by reporting the problem and somehow coping with the
389	with the watcher being stopped.	1301	watcher being stopped. Note that well-written programs should not receive
		1302	an error ever, so when your watcher receives it, this usually indicates a
		1303	bug in your program.
390		1304
391	Libev will usually signal a few "dummy" events together with an error,	1305	Libev will usually signal a few "dummy" events together with an error, for
392	for example it might indicate that a fd is readable or writable, and if	1306	example it might indicate that a fd is readable or writable, and if your
393	your callbacks is well-written it can just attempt the operation and cope	1307	callbacks is well-written it can just attempt the operation and cope with
394	with the error from read() or write(). This will not work in multithreaded	1308	the error from read() or write(). This will not work in multi-threaded
395	programs, though, so beware.	1309	programs, though, as the fd could already be closed and reused for another
		1310	thing, so beware.
396		1311
397	=back	1312	=back
398		1313
399	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER	1314	=head2 GENERIC WATCHER FUNCTIONS
400		1315
401	Each watcher has, by default, a member C<void *data> that you can change	1316	=over 4
402	and read at any time, libev will completely ignore it. This can be used
403	to associate arbitrary data with your watcher. If you need more data and
404	don't want to allocate memory and store a pointer to it in that data
405	member, you can also "subclass" the watcher type and provide your own
406	data:
407		1317
408	struct my_io	1318	=item C<ev_init> (ev_TYPE *watcher, callback)
		1319
		1320	This macro initialises the generic portion of a watcher. The contents
		1321	of the watcher object can be arbitrary (so C<malloc> will do). Only
		1322	the generic parts of the watcher are initialised, you I<need> to call
		1323	the type-specific C<ev_TYPE_set> macro afterwards to initialise the
		1324	type-specific parts. For each type there is also a C<ev_TYPE_init> macro
		1325	which rolls both calls into one.
		1326
		1327	You can reinitialise a watcher at any time as long as it has been stopped
		1328	(or never started) and there are no pending events outstanding.
		1329
		1330	The callback is always of type C<void (*)(struct ev_loop *loop, ev_TYPE *watcher,
		1331	int revents)>.
		1332
		1333	Example: Initialise an C<ev_io> watcher in two steps.
		1334
		1335	ev_io w;
		1336	ev_init (&w, my_cb);
		1337	ev_io_set (&w, STDIN_FILENO, EV_READ);
		1338
		1339	=item C<ev_TYPE_set> (ev_TYPE *watcher, [args])
		1340
		1341	This macro initialises the type-specific parts of a watcher. You need to
		1342	call C<ev_init> at least once before you call this macro, but you can
		1343	call C<ev_TYPE_set> any number of times. You must not, however, call this
		1344	macro on a watcher that is active (it can be pending, however, which is a
		1345	difference to the C<ev_init> macro).
		1346
		1347	Although some watcher types do not have type-specific arguments
		1348	(e.g. C<ev_prepare>) you still need to call its C<set> macro.
		1349
		1350	See C<ev_init>, above, for an example.
		1351
		1352	=item C<ev_TYPE_init> (ev_TYPE *watcher, callback, [args])
		1353
		1354	This convenience macro rolls both C<ev_init> and C<ev_TYPE_set> macro
		1355	calls into a single call. This is the most convenient method to initialise
		1356	a watcher. The same limitations apply, of course.
		1357
		1358	Example: Initialise and set an C<ev_io> watcher in one step.
		1359
		1360	ev_io_init (&w, my_cb, STDIN_FILENO, EV_READ);
		1361
		1362	=item C<ev_TYPE_start> (loop, ev_TYPE *watcher)
		1363
		1364	Starts (activates) the given watcher. Only active watchers will receive
		1365	events. If the watcher is already active nothing will happen.
		1366
		1367	Example: Start the C<ev_io> watcher that is being abused as example in this
		1368	whole section.
		1369
		1370	ev_io_start (EV_DEFAULT_UC, &w);
		1371
		1372	=item C<ev_TYPE_stop> (loop, ev_TYPE *watcher)
		1373
		1374	Stops the given watcher if active, and clears the pending status (whether
		1375	the watcher was active or not).
		1376
		1377	It is possible that stopped watchers are pending - for example,
		1378	non-repeating timers are being stopped when they become pending - but
		1379	calling C<ev_TYPE_stop> ensures that the watcher is neither active nor
		1380	pending. If you want to free or reuse the memory used by the watcher it is
		1381	therefore a good idea to always call its C<ev_TYPE_stop> function.
		1382
		1383	=item bool ev_is_active (ev_TYPE *watcher)
		1384
		1385	Returns a true value iff the watcher is active (i.e. it has been started
		1386	and not yet been stopped). As long as a watcher is active you must not modify
		1387	it.
		1388
		1389	=item bool ev_is_pending (ev_TYPE *watcher)
		1390
		1391	Returns a true value iff the watcher is pending, (i.e. it has outstanding
		1392	events but its callback has not yet been invoked). As long as a watcher
		1393	is pending (but not active) you must not call an init function on it (but
		1394	C<ev_TYPE_set> is safe), you must not change its priority, and you must
		1395	make sure the watcher is available to libev (e.g. you cannot C<free ()>
		1396	it).
		1397
		1398	=item callback ev_cb (ev_TYPE *watcher)
		1399
		1400	Returns the callback currently set on the watcher.
		1401
		1402	=item ev_set_cb (ev_TYPE *watcher, callback)
		1403
		1404	Change the callback. You can change the callback at virtually any time
		1405	(modulo threads).
		1406
		1407	=item ev_set_priority (ev_TYPE *watcher, int priority)
		1408
		1409	=item int ev_priority (ev_TYPE *watcher)
		1410
		1411	Set and query the priority of the watcher. The priority is a small
		1412	integer between C<EV_MAXPRI> (default: C<2>) and C<EV_MINPRI>
		1413	(default: C<-2>). Pending watchers with higher priority will be invoked
		1414	before watchers with lower priority, but priority will not keep watchers
		1415	from being executed (except for C<ev_idle> watchers).
		1416
		1417	If you need to suppress invocation when higher priority events are pending
		1418	you need to look at C<ev_idle> watchers, which provide this functionality.
		1419
		1420	You I<must not> change the priority of a watcher as long as it is active or
		1421	pending.
		1422
		1423	Setting a priority outside the range of C<EV_MINPRI> to C<EV_MAXPRI> is
		1424	fine, as long as you do not mind that the priority value you query might
		1425	or might not have been clamped to the valid range.
		1426
		1427	The default priority used by watchers when no priority has been set is
		1428	always C<0>, which is supposed to not be too high and not be too low :).
		1429
		1430	See L</WATCHER PRIORITY MODELS>, below, for a more thorough treatment of
		1431	priorities.
		1432
		1433	=item ev_invoke (loop, ev_TYPE *watcher, int revents)
		1434
		1435	Invoke the C<watcher> with the given C<loop> and C<revents>. Neither
		1436	C<loop> nor C<revents> need to be valid as long as the watcher callback
		1437	can deal with that fact, as both are simply passed through to the
		1438	callback.
		1439
		1440	=item int ev_clear_pending (loop, ev_TYPE *watcher)
		1441
		1442	If the watcher is pending, this function clears its pending status and
		1443	returns its C<revents> bitset (as if its callback was invoked). If the
		1444	watcher isn't pending it does nothing and returns C<0>.
		1445
		1446	Sometimes it can be useful to "poll" a watcher instead of waiting for its
		1447	callback to be invoked, which can be accomplished with this function.
		1448
		1449	=item ev_feed_event (loop, ev_TYPE *watcher, int revents)
		1450
		1451	Feeds the given event set into the event loop, as if the specified event
		1452	had happened for the specified watcher (which must be a pointer to an
		1453	initialised but not necessarily started event watcher). Obviously you must
		1454	not free the watcher as long as it has pending events.
		1455
		1456	Stopping the watcher, letting libev invoke it, or calling
		1457	C<ev_clear_pending> will clear the pending event, even if the watcher was
		1458	not started in the first place.
		1459
		1460	See also C<ev_feed_fd_event> and C<ev_feed_signal_event> for related
		1461	functions that do not need a watcher.
		1462
		1463	=back
		1464
		1465	See also the L</ASSOCIATING CUSTOM DATA WITH A WATCHER> and L</BUILDING YOUR
		1466	OWN COMPOSITE WATCHERS> idioms.
		1467
		1468	=head2 WATCHER STATES
		1469
		1470	There are various watcher states mentioned throughout this manual -
		1471	active, pending and so on. In this section these states and the rules to
		1472	transition between them will be described in more detail - and while these
		1473	rules might look complicated, they usually do "the right thing".
		1474
		1475	=over 4
		1476
		1477	=item initialised
		1478
		1479	Before a watcher can be registered with the event loop it has to be
		1480	initialised. This can be done with a call to C<ev_TYPE_init>, or calls to
		1481	C<ev_init> followed by the watcher-specific C<ev_TYPE_set> function.
		1482
		1483	In this state it is simply some block of memory that is suitable for
		1484	use in an event loop. It can be moved around, freed, reused etc. at
		1485	will - as long as you either keep the memory contents intact, or call
		1486	C<ev_TYPE_init> again.
		1487
		1488	=item started/running/active
		1489
		1490	Once a watcher has been started with a call to C<ev_TYPE_start> it becomes
		1491	property of the event loop, and is actively waiting for events. While in
		1492	this state it cannot be accessed (except in a few documented ways), moved,
		1493	freed or anything else - the only legal thing is to keep a pointer to it,
		1494	and call libev functions on it that are documented to work on active watchers.
		1495
		1496	=item pending
		1497
		1498	If a watcher is active and libev determines that an event it is interested
		1499	in has occurred (such as a timer expiring), it will become pending. It will
		1500	stay in this pending state until either it is stopped or its callback is
		1501	about to be invoked, so it is not normally pending inside the watcher
		1502	callback.
		1503
		1504	The watcher might or might not be active while it is pending (for example,
		1505	an expired non-repeating timer can be pending but no longer active). If it
		1506	is stopped, it can be freely accessed (e.g. by calling C<ev_TYPE_set>),
		1507	but it is still property of the event loop at this time, so cannot be
		1508	moved, freed or reused. And if it is active the rules described in the
		1509	previous item still apply.
		1510
		1511	It is also possible to feed an event on a watcher that is not active (e.g.
		1512	via C<ev_feed_event>), in which case it becomes pending without being
		1513	active.
		1514
		1515	=item stopped
		1516
		1517	A watcher can be stopped implicitly by libev (in which case it might still
		1518	be pending), or explicitly by calling its C<ev_TYPE_stop> function. The
		1519	latter will clear any pending state the watcher might be in, regardless
		1520	of whether it was active or not, so stopping a watcher explicitly before
		1521	freeing it is often a good idea.
		1522
		1523	While stopped (and not pending) the watcher is essentially in the
		1524	initialised state, that is, it can be reused, moved, modified in any way
		1525	you wish (but when you trash the memory block, you need to C<ev_TYPE_init>
		1526	it again).
		1527
		1528	=back
		1529
		1530	=head2 WATCHER PRIORITY MODELS
		1531
		1532	Many event loops support I<watcher priorities>, which are usually small
		1533	integers that influence the ordering of event callback invocation
		1534	between watchers in some way, all else being equal.
		1535
		1536	In libev, Watcher priorities can be set using C<ev_set_priority>. See its
		1537	description for the more technical details such as the actual priority
		1538	range.
		1539
		1540	There are two common ways how these these priorities are being interpreted
		1541	by event loops:
		1542
		1543	In the more common lock-out model, higher priorities "lock out" invocation
		1544	of lower priority watchers, which means as long as higher priority
		1545	watchers receive events, lower priority watchers are not being invoked.
		1546
		1547	The less common only-for-ordering model uses priorities solely to order
		1548	callback invocation within a single event loop iteration: Higher priority
		1549	watchers are invoked before lower priority ones, but they all get invoked
		1550	before polling for new events.
		1551
		1552	Libev uses the second (only-for-ordering) model for all its watchers
		1553	except for idle watchers (which use the lock-out model).
		1554
		1555	The rationale behind this is that implementing the lock-out model for
		1556	watchers is not well supported by most kernel interfaces, and most event
		1557	libraries will just poll for the same events again and again as long as
		1558	their callbacks have not been executed, which is very inefficient in the
		1559	common case of one high-priority watcher locking out a mass of lower
		1560	priority ones.
		1561
		1562	Static (ordering) priorities are most useful when you have two or more
		1563	watchers handling the same resource: a typical usage example is having an
		1564	C<ev_io> watcher to receive data, and an associated C<ev_timer> to handle
		1565	timeouts. Under load, data might be received while the program handles
		1566	other jobs, but since timers normally get invoked first, the timeout
		1567	handler will be executed before checking for data. In that case, giving
		1568	the timer a lower priority than the I/O watcher ensures that I/O will be
		1569	handled first even under adverse conditions (which is usually, but not
		1570	always, what you want).
		1571
		1572	Since idle watchers use the "lock-out" model, meaning that idle watchers
		1573	will only be executed when no same or higher priority watchers have
		1574	received events, they can be used to implement the "lock-out" model when
		1575	required.
		1576
		1577	For example, to emulate how many other event libraries handle priorities,
		1578	you can associate an C<ev_idle> watcher to each such watcher, and in
		1579	the normal watcher callback, you just start the idle watcher. The real
		1580	processing is done in the idle watcher callback. This causes libev to
		1581	continuously poll and process kernel event data for the watcher, but when
		1582	the lock-out case is known to be rare (which in turn is rare :), this is
		1583	workable.
		1584
		1585	Usually, however, the lock-out model implemented that way will perform
		1586	miserably under the type of load it was designed to handle. In that case,
		1587	it might be preferable to stop the real watcher before starting the
		1588	idle watcher, so the kernel will not have to process the event in case
		1589	the actual processing will be delayed for considerable time.
		1590
		1591	Here is an example of an I/O watcher that should run at a strictly lower
		1592	priority than the default, and which should only process data when no
		1593	other events are pending:
		1594
		1595	ev_idle idle; // actual processing watcher
		1596	ev_io io; // actual event watcher
		1597
		1598	static void
		1599	io_cb (EV_P_ ev_io *w, int revents)
409	{	1600	{
410	struct ev_io io;	1601	// stop the I/O watcher, we received the event, but
411	int otherfd;	1602	// are not yet ready to handle it.
412	void *somedata;	1603	ev_io_stop (EV_A_ w);
413	struct whatever *mostinteresting;	1604
		1605	// start the idle watcher to handle the actual event.
		1606	// it will not be executed as long as other watchers
		1607	// with the default priority are receiving events.
		1608	ev_idle_start (EV_A_ &idle);
414	}	1609	}
415		1610
416	And since your callback will be called with a pointer to the watcher, you	1611	static void
417	can cast it back to your own type:	1612	idle_cb (EV_P_ ev_idle *w, int revents)
418
419	static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)
420	{	1613	{
421	struct my_io *w = (struct my_io *)w_;	1614	// actual processing
422	...	1615	read (STDIN_FILENO, ...);
		1616
		1617	// have to start the I/O watcher again, as
		1618	// we have handled the event
		1619	ev_io_start (EV_P_ &io);
423	}	1620	}
424		1621
425	More interesting and less C-conformant ways of catsing your callback type	1622	// initialisation
426	have been omitted....	1623	ev_idle_init (&idle, idle_cb);
		1624	ev_io_init (&io, io_cb, STDIN_FILENO, EV_READ);
		1625	ev_io_start (EV_DEFAULT_ &io);
		1626
		1627	In the "real" world, it might also be beneficial to start a timer, so that
		1628	low-priority connections can not be locked out forever under load. This
		1629	enables your program to keep a lower latency for important connections
		1630	during short periods of high load, while not completely locking out less
		1631	important ones.
427		1632
428		1633
429	=head1 WATCHER TYPES	1634	=head1 WATCHER TYPES
430		1635
431	This section describes each watcher in detail, but will not repeat	1636	This section describes each watcher in detail, but will not repeat
432	information given in the last section.	1637	information given in the last section. Any initialisation/set macros,
		1638	functions and members specific to the watcher type are explained.
433		1639
		1640	Members are additionally marked with either I<[read-only]>, meaning that,
		1641	while the watcher is active, you can look at the member and expect some
		1642	sensible content, but you must not modify it (you can modify it while the
		1643	watcher is stopped to your hearts content), or I<[read-write]>, which
		1644	means you can expect it to have some sensible content while the watcher
		1645	is active, but you can also modify it. Modifying it may not do something
		1646	sensible or take immediate effect (or do anything at all), but libev will
		1647	not crash or malfunction in any way.
		1648
		1649
434	=head2 C<ev_io> - is this file descriptor readable or writable	1650	=head2 C<ev_io> - is this file descriptor readable or writable?
435		1651
436	I/O watchers check whether a file descriptor is readable or writable	1652	I/O watchers check whether a file descriptor is readable or writable
437	in each iteration of the event loop (This behaviour is called	1653	in each iteration of the event loop, or, more precisely, when reading
438	level-triggering because you keep receiving events as long as the	1654	would not block the process and writing would at least be able to write
439	condition persists. Remember you can stop the watcher if you don't want to	1655	some data. This behaviour is called level-triggering because you keep
440	act on the event and neither want to receive future events).	1656	receiving events as long as the condition persists. Remember you can stop
		1657	the watcher if you don't want to act on the event and neither want to
		1658	receive future events.
441		1659
442	In general you can register as many read and/or write event watchers per	1660	In general you can register as many read and/or write event watchers per
443	fd as you want (as long as you don't confuse yourself). Setting all file	1661	fd as you want (as long as you don't confuse yourself). Setting all file
444	descriptors to non-blocking mode is also usually a good idea (but not	1662	descriptors to non-blocking mode is also usually a good idea (but not
445	required if you know what you are doing).	1663	required if you know what you are doing).
446		1664
447	You have to be careful with dup'ed file descriptors, though. Some backends	1665	Another thing you have to watch out for is that it is quite easy to
448	(the linux epoll backend is a notable example) cannot handle dup'ed file	1666	receive "spurious" readiness notifications, that is, your callback might
449	descriptors correctly if you register interest in two or more fds pointing	1667	be called with C<EV_READ> but a subsequent C<read>(2) will actually block
450	to the same underlying file/socket etc. description (that is, they share	1668	because there is no data. It is very easy to get into this situation even
451	the same underlying "file open").	1669	with a relatively standard program structure. Thus it is best to always
		1670	use non-blocking I/O: An extra C<read>(2) returning C<EAGAIN> is far
		1671	preferable to a program hanging until some data arrives.
452		1672
453	If you must do this, then force the use of a known-to-be-good backend	1673	If you cannot run the fd in non-blocking mode (for example you should
454	(at the time of this writing, this includes only EVMETHOD_SELECT and	1674	not play around with an Xlib connection), then you have to separately
455	EVMETHOD_POLL).	1675	re-test whether a file descriptor is really ready with a known-to-be good
		1676	interface such as poll (fortunately in the case of Xlib, it already does
		1677	this on its own, so its quite safe to use). Some people additionally
		1678	use C<SIGALRM> and an interval timer, just to be sure you won't block
		1679	indefinitely.
		1680
		1681	But really, best use non-blocking mode.
		1682
		1683	=head3 The special problem of disappearing file descriptors
		1684
		1685	Some backends (e.g. kqueue, epoll, linuxaio) need to be told about closing
		1686	a file descriptor (either due to calling C<close> explicitly or any other
		1687	means, such as C<dup2>). The reason is that you register interest in some
		1688	file descriptor, but when it goes away, the operating system will silently
		1689	drop this interest. If another file descriptor with the same number then
		1690	is registered with libev, there is no efficient way to see that this is,
		1691	in fact, a different file descriptor.
		1692
		1693	To avoid having to explicitly tell libev about such cases, libev follows
		1694	the following policy: Each time C<ev_io_set> is being called, libev
		1695	will assume that this is potentially a new file descriptor, otherwise
		1696	it is assumed that the file descriptor stays the same. That means that
		1697	you I<have> to call C<ev_io_set> (or C<ev_io_init>) when you change the
		1698	descriptor even if the file descriptor number itself did not change.
		1699
		1700	This is how one would do it normally anyway, the important point is that
		1701	the libev application should not optimise around libev but should leave
		1702	optimisations to libev.
		1703
		1704	=head3 The special problem of dup'ed file descriptors
		1705
		1706	Some backends (e.g. epoll), cannot register events for file descriptors,
		1707	but only events for the underlying file descriptions. That means when you
		1708	have C<dup ()>'ed file descriptors or weirder constellations, and register
		1709	events for them, only one file descriptor might actually receive events.
		1710
		1711	There is no workaround possible except not registering events
		1712	for potentially C<dup ()>'ed file descriptors, or to resort to
		1713	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1714
		1715	=head3 The special problem of files
		1716
		1717	Many people try to use C<select> (or libev) on file descriptors
		1718	representing files, and expect it to become ready when their program
		1719	doesn't block on disk accesses (which can take a long time on their own).
		1720
		1721	However, this cannot ever work in the "expected" way - you get a readiness
		1722	notification as soon as the kernel knows whether and how much data is
		1723	there, and in the case of open files, that's always the case, so you
		1724	always get a readiness notification instantly, and your read (or possibly
		1725	write) will still block on the disk I/O.
		1726
		1727	Another way to view it is that in the case of sockets, pipes, character
		1728	devices and so on, there is another party (the sender) that delivers data
		1729	on its own, but in the case of files, there is no such thing: the disk
		1730	will not send data on its own, simply because it doesn't know what you
		1731	wish to read - you would first have to request some data.
		1732
		1733	Since files are typically not-so-well supported by advanced notification
		1734	mechanism, libev tries hard to emulate POSIX behaviour with respect
		1735	to files, even though you should not use it. The reason for this is
		1736	convenience: sometimes you want to watch STDIN or STDOUT, which is
		1737	usually a tty, often a pipe, but also sometimes files or special devices
		1738	(for example, C<epoll> on Linux works with F</dev/random> but not with
		1739	F</dev/urandom>), and even though the file might better be served with
		1740	asynchronous I/O instead of with non-blocking I/O, it is still useful when
		1741	it "just works" instead of freezing.
		1742
		1743	So avoid file descriptors pointing to files when you know it (e.g. use
		1744	libeio), but use them when it is convenient, e.g. for STDIN/STDOUT, or
		1745	when you rarely read from a file instead of from a socket, and want to
		1746	reuse the same code path.
		1747
		1748	=head3 The special problem of fork
		1749
		1750	Some backends (epoll, kqueue, probably linuxaio) do not support C<fork ()>
		1751	at all or exhibit useless behaviour. Libev fully supports fork, but needs
		1752	to be told about it in the child if you want to continue to use it in the
		1753	child.
		1754
		1755	To support fork in your child processes, you have to call C<ev_loop_fork
		1756	()> after a fork in the child, enable C<EVFLAG_FORKCHECK>, or resort to
		1757	C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>.
		1758
		1759	=head3 The special problem of SIGPIPE
		1760
		1761	While not really specific to libev, it is easy to forget about C<SIGPIPE>:
		1762	when writing to a pipe whose other end has been closed, your program gets
		1763	sent a SIGPIPE, which, by default, aborts your program. For most programs
		1764	this is sensible behaviour, for daemons, this is usually undesirable.
		1765
		1766	So when you encounter spurious, unexplained daemon exits, make sure you
		1767	ignore SIGPIPE (and maybe make sure you log the exit status of your daemon
		1768	somewhere, as that would have given you a big clue).
		1769
		1770	=head3 The special problem of accept()ing when you can't
		1771
		1772	Many implementations of the POSIX C<accept> function (for example,
		1773	found in post-2004 Linux) have the peculiar behaviour of not removing a
		1774	connection from the pending queue in all error cases.
		1775
		1776	For example, larger servers often run out of file descriptors (because
		1777	of resource limits), causing C<accept> to fail with C<ENFILE> but not
		1778	rejecting the connection, leading to libev signalling readiness on
		1779	the next iteration again (the connection still exists after all), and
		1780	typically causing the program to loop at 100% CPU usage.
		1781
		1782	Unfortunately, the set of errors that cause this issue differs between
		1783	operating systems, there is usually little the app can do to remedy the
		1784	situation, and no known thread-safe method of removing the connection to
		1785	cope with overload is known (to me).
		1786
		1787	One of the easiest ways to handle this situation is to just ignore it
		1788	- when the program encounters an overload, it will just loop until the
		1789	situation is over. While this is a form of busy waiting, no OS offers an
		1790	event-based way to handle this situation, so it's the best one can do.
		1791
		1792	A better way to handle the situation is to log any errors other than
		1793	C<EAGAIN> and C<EWOULDBLOCK>, making sure not to flood the log with such
		1794	messages, and continue as usual, which at least gives the user an idea of
		1795	what could be wrong ("raise the ulimit!"). For extra points one could stop
		1796	the C<ev_io> watcher on the listening fd "for a while", which reduces CPU
		1797	usage.
		1798
		1799	If your program is single-threaded, then you could also keep a dummy file
		1800	descriptor for overload situations (e.g. by opening F</dev/null>), and
		1801	when you run into C<ENFILE> or C<EMFILE>, close it, run C<accept>,
		1802	close that fd, and create a new dummy fd. This will gracefully refuse
		1803	clients under typical overload conditions.
		1804
		1805	The last way to handle it is to simply log the error and C<exit>, as
		1806	is often done with C<malloc> failures, but this results in an easy
		1807	opportunity for a DoS attack.
		1808
		1809	=head3 Watcher-Specific Functions
456		1810
457	=over 4	1811	=over 4
458		1812
459	=item ev_io_init (ev_io *, callback, int fd, int events)	1813	=item ev_io_init (ev_io *, callback, int fd, int events)
460		1814
461	=item ev_io_set (ev_io *, int fd, int events)	1815	=item ev_io_set (ev_io *, int fd, int events)
462		1816
463	Configures an C<ev_io> watcher. The fd is the file descriptor to rceeive	1817	Configures an C<ev_io> watcher. The C<fd> is the file descriptor to
464	events for and events is either C<EV_READ>, C<EV_WRITE> or C<EV_READ |	1818	receive events for and C<events> is either C<EV_READ>, C<EV_WRITE> or
465	EV_WRITE> to receive the given events.	1819	C<EV_READ | EV_WRITE>, to express the desire to receive the given events.
		1820
		1821	=item int fd [read-only]
		1822
		1823	The file descriptor being watched.
		1824
		1825	=item int events [read-only]
		1826
		1827	The events being watched.
466		1828
467	=back	1829	=back
468		1830
		1831	=head3 Examples
		1832
		1833	Example: Call C<stdin_readable_cb> when STDIN_FILENO has become, well
		1834	readable, but only once. Since it is likely line-buffered, you could
		1835	attempt to read a whole line in the callback.
		1836
		1837	static void
		1838	stdin_readable_cb (struct ev_loop *loop, ev_io *w, int revents)
		1839	{
		1840	ev_io_stop (loop, w);
		1841	.. read from stdin here (or from w->fd) and handle any I/O errors
		1842	}
		1843
		1844	...
		1845	struct ev_loop *loop = ev_default_init (0);
		1846	ev_io stdin_readable;
		1847	ev_io_init (&stdin_readable, stdin_readable_cb, STDIN_FILENO, EV_READ);
		1848	ev_io_start (loop, &stdin_readable);
		1849	ev_run (loop, 0);
		1850
		1851
469	=head2 C<ev_timer> - relative and optionally recurring timeouts	1852	=head2 C<ev_timer> - relative and optionally repeating timeouts
470		1853
471	Timer watchers are simple relative timers that generate an event after a	1854	Timer watchers are simple relative timers that generate an event after a
472	given time, and optionally repeating in regular intervals after that.	1855	given time, and optionally repeating in regular intervals after that.
473		1856
474	The timers are based on real time, that is, if you register an event that	1857	The timers are based on real time, that is, if you register an event that
475	times out after an hour and you reset your system clock to last years	1858	times out after an hour and you reset your system clock to January last
476	time, it will still time out after (roughly) and hour. "Roughly" because	1859	year, it will still time out after (roughly) one hour. "Roughly" because
477	detecting time jumps is hard, and soem inaccuracies are unavoidable (the	1860	detecting time jumps is hard, and some inaccuracies are unavoidable (the
478	monotonic clock option helps a lot here).	1861	monotonic clock option helps a lot here).
		1862
		1863	The callback is guaranteed to be invoked only I<after> its timeout has
		1864	passed (not I<at>, so on systems with very low-resolution clocks this
		1865	might introduce a small delay, see "the special problem of being too
		1866	early", below). If multiple timers become ready during the same loop
		1867	iteration then the ones with earlier time-out values are invoked before
		1868	ones of the same priority with later time-out values (but this is no
		1869	longer true when a callback calls C<ev_run> recursively).
		1870
		1871	=head3 Be smart about timeouts
		1872
		1873	Many real-world problems involve some kind of timeout, usually for error
		1874	recovery. A typical example is an HTTP request - if the other side hangs,
		1875	you want to raise some error after a while.
		1876
		1877	What follows are some ways to handle this problem, from obvious and
		1878	inefficient to smart and efficient.
		1879
		1880	In the following, a 60 second activity timeout is assumed - a timeout that
		1881	gets reset to 60 seconds each time there is activity (e.g. each time some
		1882	data or other life sign was received).
		1883
		1884	=over 4
		1885
		1886	=item 1. Use a timer and stop, reinitialise and start it on activity.
		1887
		1888	This is the most obvious, but not the most simple way: In the beginning,
		1889	start the watcher:
		1890
		1891	ev_timer_init (timer, callback, 60., 0.);
		1892	ev_timer_start (loop, timer);
		1893
		1894	Then, each time there is some activity, C<ev_timer_stop> it, initialise it
		1895	and start it again:
		1896
		1897	ev_timer_stop (loop, timer);
		1898	ev_timer_set (timer, 60., 0.);
		1899	ev_timer_start (loop, timer);
		1900
		1901	This is relatively simple to implement, but means that each time there is
		1902	some activity, libev will first have to remove the timer from its internal
		1903	data structure and then add it again. Libev tries to be fast, but it's
		1904	still not a constant-time operation.
		1905
		1906	=item 2. Use a timer and re-start it with C<ev_timer_again> inactivity.
		1907
		1908	This is the easiest way, and involves using C<ev_timer_again> instead of
		1909	C<ev_timer_start>.
		1910
		1911	To implement this, configure an C<ev_timer> with a C<repeat> value
		1912	of C<60> and then call C<ev_timer_again> at start and each time you
		1913	successfully read or write some data. If you go into an idle state where
		1914	you do not expect data to travel on the socket, you can C<ev_timer_stop>
		1915	the timer, and C<ev_timer_again> will automatically restart it if need be.
		1916
		1917	That means you can ignore both the C<ev_timer_start> function and the
		1918	C<after> argument to C<ev_timer_set>, and only ever use the C<repeat>
		1919	member and C<ev_timer_again>.
		1920
		1921	At start:
		1922
		1923	ev_init (timer, callback);
		1924	timer->repeat = 60.;
		1925	ev_timer_again (loop, timer);
		1926
		1927	Each time there is some activity:
		1928
		1929	ev_timer_again (loop, timer);
		1930
		1931	It is even possible to change the time-out on the fly, regardless of
		1932	whether the watcher is active or not:
		1933
		1934	timer->repeat = 30.;
		1935	ev_timer_again (loop, timer);
		1936
		1937	This is slightly more efficient then stopping/starting the timer each time
		1938	you want to modify its timeout value, as libev does not have to completely
		1939	remove and re-insert the timer from/into its internal data structure.
		1940
		1941	It is, however, even simpler than the "obvious" way to do it.
		1942
		1943	=item 3. Let the timer time out, but then re-arm it as required.
		1944
		1945	This method is more tricky, but usually most efficient: Most timeouts are
		1946	relatively long compared to the intervals between other activity - in
		1947	our example, within 60 seconds, there are usually many I/O events with
		1948	associated activity resets.
		1949
		1950	In this case, it would be more efficient to leave the C<ev_timer> alone,
		1951	but remember the time of last activity, and check for a real timeout only
		1952	within the callback:
		1953
		1954	ev_tstamp timeout = 60.;
		1955	ev_tstamp last_activity; // time of last activity
		1956	ev_timer timer;
		1957
		1958	static void
		1959	callback (EV_P_ ev_timer *w, int revents)
		1960	{
		1961	// calculate when the timeout would happen
		1962	ev_tstamp after = last_activity - ev_now (EV_A) + timeout;
		1963
		1964	// if negative, it means we the timeout already occurred
		1965	if (after < 0.)
		1966	{
		1967	// timeout occurred, take action
		1968	}
		1969	else
		1970	{
		1971	// callback was invoked, but there was some recent
		1972	// activity. simply restart the timer to time out
		1973	// after "after" seconds, which is the earliest time
		1974	// the timeout can occur.
		1975	ev_timer_set (w, after, 0.);
		1976	ev_timer_start (EV_A_ w);
		1977	}
		1978	}
		1979
		1980	To summarise the callback: first calculate in how many seconds the
		1981	timeout will occur (by calculating the absolute time when it would occur,
		1982	C<last_activity + timeout>, and subtracting the current time, C<ev_now
		1983	(EV_A)> from that).
		1984
		1985	If this value is negative, then we are already past the timeout, i.e. we
		1986	timed out, and need to do whatever is needed in this case.
		1987
		1988	Otherwise, we now the earliest time at which the timeout would trigger,
		1989	and simply start the timer with this timeout value.
		1990
		1991	In other words, each time the callback is invoked it will check whether
		1992	the timeout occurred. If not, it will simply reschedule itself to check
		1993	again at the earliest time it could time out. Rinse. Repeat.
		1994
		1995	This scheme causes more callback invocations (about one every 60 seconds
		1996	minus half the average time between activity), but virtually no calls to
		1997	libev to change the timeout.
		1998
		1999	To start the machinery, simply initialise the watcher and set
		2000	C<last_activity> to the current time (meaning there was some activity just
		2001	now), then call the callback, which will "do the right thing" and start
		2002	the timer:
		2003
		2004	last_activity = ev_now (EV_A);
		2005	ev_init (&timer, callback);
		2006	callback (EV_A_ &timer, 0);
		2007
		2008	When there is some activity, simply store the current time in
		2009	C<last_activity>, no libev calls at all:
		2010
		2011	if (activity detected)
		2012	last_activity = ev_now (EV_A);
		2013
		2014	When your timeout value changes, then the timeout can be changed by simply
		2015	providing a new value, stopping the timer and calling the callback, which
		2016	will again do the right thing (for example, time out immediately :).
		2017
		2018	timeout = new_value;
		2019	ev_timer_stop (EV_A_ &timer);
		2020	callback (EV_A_ &timer, 0);
		2021
		2022	This technique is slightly more complex, but in most cases where the
		2023	time-out is unlikely to be triggered, much more efficient.
		2024
		2025	=item 4. Wee, just use a double-linked list for your timeouts.
		2026
		2027	If there is not one request, but many thousands (millions...), all
		2028	employing some kind of timeout with the same timeout value, then one can
		2029	do even better:
		2030
		2031	When starting the timeout, calculate the timeout value and put the timeout
		2032	at the I<end> of the list.
		2033
		2034	Then use an C<ev_timer> to fire when the timeout at the I<beginning> of
		2035	the list is expected to fire (for example, using the technique #3).
		2036
		2037	When there is some activity, remove the timer from the list, recalculate
		2038	the timeout, append it to the end of the list again, and make sure to
		2039	update the C<ev_timer> if it was taken from the beginning of the list.
		2040
		2041	This way, one can manage an unlimited number of timeouts in O(1) time for
		2042	starting, stopping and updating the timers, at the expense of a major
		2043	complication, and having to use a constant timeout. The constant timeout
		2044	ensures that the list stays sorted.
		2045
		2046	=back
		2047
		2048	So which method the best?
		2049
		2050	Method #2 is a simple no-brain-required solution that is adequate in most
		2051	situations. Method #3 requires a bit more thinking, but handles many cases
		2052	better, and isn't very complicated either. In most case, choosing either
		2053	one is fine, with #3 being better in typical situations.
		2054
		2055	Method #1 is almost always a bad idea, and buys you nothing. Method #4 is
		2056	rather complicated, but extremely efficient, something that really pays
		2057	off after the first million or so of active timers, i.e. it's usually
		2058	overkill :)
		2059
		2060	=head3 The special problem of being too early
		2061
		2062	If you ask a timer to call your callback after three seconds, then
		2063	you expect it to be invoked after three seconds - but of course, this
		2064	cannot be guaranteed to infinite precision. Less obviously, it cannot be
		2065	guaranteed to any precision by libev - imagine somebody suspending the
		2066	process with a STOP signal for a few hours for example.
		2067
		2068	So, libev tries to invoke your callback as soon as possible I<after> the
		2069	delay has occurred, but cannot guarantee this.
		2070
		2071	A less obvious failure mode is calling your callback too early: many event
		2072	loops compare timestamps with a "elapsed delay >= requested delay", but
		2073	this can cause your callback to be invoked much earlier than you would
		2074	expect.
		2075
		2076	To see why, imagine a system with a clock that only offers full second
		2077	resolution (think windows if you can't come up with a broken enough OS
		2078	yourself). If you schedule a one-second timer at the time 500.9, then the
		2079	event loop will schedule your timeout to elapse at a system time of 500
		2080	(500.9 truncated to the resolution) + 1, or 501.
		2081
		2082	If an event library looks at the timeout 0.1s later, it will see "501 >=
		2083	501" and invoke the callback 0.1s after it was started, even though a
		2084	one-second delay was requested - this is being "too early", despite best
		2085	intentions.
		2086
		2087	This is the reason why libev will never invoke the callback if the elapsed
		2088	delay equals the requested delay, but only when the elapsed delay is
		2089	larger than the requested delay. In the example above, libev would only invoke
		2090	the callback at system time 502, or 1.1s after the timer was started.
		2091
		2092	So, while libev cannot guarantee that your callback will be invoked
		2093	exactly when requested, it I<can> and I<does> guarantee that the requested
		2094	delay has actually elapsed, or in other words, it always errs on the "too
		2095	late" side of things.
		2096
		2097	=head3 The special problem of time updates
		2098
		2099	Establishing the current time is a costly operation (it usually takes
		2100	at least one system call): EV therefore updates its idea of the current
		2101	time only before and after C<ev_run> collects new events, which causes a
		2102	growing difference between C<ev_now ()> and C<ev_time ()> when handling
		2103	lots of events in one iteration.
479		2104
480	The relative timeouts are calculated relative to the C<ev_now ()>	2105	The relative timeouts are calculated relative to the C<ev_now ()>
481	time. This is usually the right thing as this timestamp refers to the time	2106	time. This is usually the right thing as this timestamp refers to the time
482	of the event triggering whatever timeout you are modifying/starting. If	2107	of the event triggering whatever timeout you are modifying/starting. If
483	you suspect event processing to be delayed and you *need* to base the timeout	2108	you suspect event processing to be delayed and you I<need> to base the
484	on the current time, use something like this to adjust for this:	2109	timeout on the current time, use something like the following to adjust
		2110	for it:
485		2111
486	ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);	2112	ev_timer_set (&timer, after + (ev_time () - ev_now ()), 0.);
		2113
		2114	If the event loop is suspended for a long time, you can also force an
		2115	update of the time returned by C<ev_now ()> by calling C<ev_now_update
		2116	()>, although that will push the event time of all outstanding events
		2117	further into the future.
		2118
		2119	=head3 The special problem of unsynchronised clocks
		2120
		2121	Modern systems have a variety of clocks - libev itself uses the normal
		2122	"wall clock" clock and, if available, the monotonic clock (to avoid time
		2123	jumps).
		2124
		2125	Neither of these clocks is synchronised with each other or any other clock
		2126	on the system, so C<ev_time ()> might return a considerably different time
		2127	than C<gettimeofday ()> or C<time ()>. On a GNU/Linux system, for example,
		2128	a call to C<gettimeofday> might return a second count that is one higher
		2129	than a directly following call to C<time>.
		2130
		2131	The moral of this is to only compare libev-related timestamps with
		2132	C<ev_time ()> and C<ev_now ()>, at least if you want better precision than
		2133	a second or so.
		2134
		2135	One more problem arises due to this lack of synchronisation: if libev uses
		2136	the system monotonic clock and you compare timestamps from C<ev_time>
		2137	or C<ev_now> from when you started your timer and when your callback is
		2138	invoked, you will find that sometimes the callback is a bit "early".
		2139
		2140	This is because C<ev_timer>s work in real time, not wall clock time, so
		2141	libev makes sure your callback is not invoked before the delay happened,
		2142	I<measured according to the real time>, not the system clock.
		2143
		2144	If your timeouts are based on a physical timescale (e.g. "time out this
		2145	connection after 100 seconds") then this shouldn't bother you as it is
		2146	exactly the right behaviour.
		2147
		2148	If you want to compare wall clock/system timestamps to your timers, then
		2149	you need to use C<ev_periodic>s, as these are based on the wall clock
		2150	time, where your comparisons will always generate correct results.
		2151
		2152	=head3 The special problems of suspended animation
		2153
		2154	When you leave the server world it is quite customary to hit machines that
		2155	can suspend/hibernate - what happens to the clocks during such a suspend?
		2156
		2157	Some quick tests made with a Linux 2.6.28 indicate that a suspend freezes
		2158	all processes, while the clocks (C<times>, C<CLOCK_MONOTONIC>) continue
		2159	to run until the system is suspended, but they will not advance while the
		2160	system is suspended. That means, on resume, it will be as if the program
		2161	was frozen for a few seconds, but the suspend time will not be counted
		2162	towards C<ev_timer> when a monotonic clock source is used. The real time
		2163	clock advanced as expected, but if it is used as sole clocksource, then a
		2164	long suspend would be detected as a time jump by libev, and timers would
		2165	be adjusted accordingly.
		2166
		2167	I would not be surprised to see different behaviour in different between
		2168	operating systems, OS versions or even different hardware.
		2169
		2170	The other form of suspend (job control, or sending a SIGSTOP) will see a
		2171	time jump in the monotonic clocks and the realtime clock. If the program
		2172	is suspended for a very long time, and monotonic clock sources are in use,
		2173	then you can expect C<ev_timer>s to expire as the full suspension time
		2174	will be counted towards the timers. When no monotonic clock source is in
		2175	use, then libev will again assume a timejump and adjust accordingly.
		2176
		2177	It might be beneficial for this latter case to call C<ev_suspend>
		2178	and C<ev_resume> in code that handles C<SIGTSTP>, to at least get
		2179	deterministic behaviour in this case (you can do nothing against
		2180	C<SIGSTOP>).
		2181
		2182	=head3 Watcher-Specific Functions and Data Members
487		2183
488	=over 4	2184	=over 4
489		2185
490	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)	2186	=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)
491		2187
492	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)	2188	=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)
493		2189
494	Configure the timer to trigger after C<after> seconds. If C<repeat> is	2190	Configure the timer to trigger after C<after> seconds (fractional and
495	C<0.>, then it will automatically be stopped. If it is positive, then the	2191	negative values are supported). If C<repeat> is C<0.>, then it will
		2192	automatically be stopped once the timeout is reached. If it is positive,
496	timer will automatically be configured to trigger again C<repeat> seconds	2193	then the timer will automatically be configured to trigger again C<repeat>
497	later, again, and again, until stopped manually.	2194	seconds later, again, and again, until stopped manually.
498		2195
499	The timer itself will do a best-effort at avoiding drift, that is, if you	2196	The timer itself will do a best-effort at avoiding drift, that is, if
500	configure a timer to trigger every 10 seconds, then it will trigger at	2197	you configure a timer to trigger every 10 seconds, then it will normally
501	exactly 10 second intervals. If, however, your program cannot keep up with	2198	trigger at exactly 10 second intervals. If, however, your program cannot
502	the timer (because it takes longer than those 10 seconds to do stuff) the	2199	keep up with the timer (because it takes longer than those 10 seconds to
503	timer will not fire more than once per event loop iteration.	2200	do stuff) the timer will not fire more than once per event loop iteration.
504		2201
505	=item ev_timer_again (loop)	2202	=item ev_timer_again (loop, ev_timer *)
506		2203
507	This will act as if the timer timed out and restart it again if it is	2204	This will act as if the timer timed out, and restarts it again if it is
508	repeating. The exact semantics are:	2205	repeating. It basically works like calling C<ev_timer_stop>, updating the
		2206	timeout to the C<repeat> value and calling C<ev_timer_start>.
509		2207
		2208	The exact semantics are as in the following rules, all of which will be
		2209	applied to the watcher:
		2210
		2211	=over 4
		2212
		2213	=item If the timer is pending, the pending status is always cleared.
		2214
510	If the timer is started but nonrepeating, stop it.	2215	=item If the timer is started but non-repeating, stop it (as if it timed
		2216	out, without invoking it).
511		2217
512	If the timer is repeating, either start it if necessary (with the repeat	2218	=item If the timer is repeating, make the C<repeat> value the new timeout
513	value), or reset the running timer to the repeat value.	2219	and start the timer, if necessary.
514
515	This sounds a bit complicated, but here is a useful and typical
516	example: Imagine you have a tcp connection and you want a so-called idle
517	timeout, that is, you want to be called when there have been, say, 60
518	seconds of inactivity on the socket. The easiest way to do this is to
519	configure an C<ev_timer> with after=repeat=60 and calling ev_timer_again each
520	time you successfully read or write some data. If you go into an idle
521	state where you do not expect data to travel on the socket, you can stop
522	the timer, and again will automatically restart it if need be.
523		2220
524	=back	2221	=back
525		2222
		2223	This sounds a bit complicated, see L</Be smart about timeouts>, above, for a
		2224	usage example.
		2225
		2226	=item ev_tstamp ev_timer_remaining (loop, ev_timer *)
		2227
		2228	Returns the remaining time until a timer fires. If the timer is active,
		2229	then this time is relative to the current event loop time, otherwise it's
		2230	the timeout value currently configured.
		2231
		2232	That is, after an C<ev_timer_set (w, 5, 7)>, C<ev_timer_remaining> returns
		2233	C<5>. When the timer is started and one second passes, C<ev_timer_remaining>
		2234	will return C<4>. When the timer expires and is restarted, it will return
		2235	roughly C<7> (likely slightly less as callback invocation takes some time,
		2236	too), and so on.
		2237
		2238	=item ev_tstamp repeat [read-write]
		2239
		2240	The current C<repeat> value. Will be used each time the watcher times out
		2241	or C<ev_timer_again> is called, and determines the next timeout (if any),
		2242	which is also when any modifications are taken into account.
		2243
		2244	=back
		2245
		2246	=head3 Examples
		2247
		2248	Example: Create a timer that fires after 60 seconds.
		2249
		2250	static void
		2251	one_minute_cb (struct ev_loop *loop, ev_timer *w, int revents)
		2252	{
		2253	.. one minute over, w is actually stopped right here
		2254	}
		2255
		2256	ev_timer mytimer;
		2257	ev_timer_init (&mytimer, one_minute_cb, 60., 0.);
		2258	ev_timer_start (loop, &mytimer);
		2259
		2260	Example: Create a timeout timer that times out after 10 seconds of
		2261	inactivity.
		2262
		2263	static void
		2264	timeout_cb (struct ev_loop *loop, ev_timer *w, int revents)
		2265	{
		2266	.. ten seconds without any activity
		2267	}
		2268
		2269	ev_timer mytimer;
		2270	ev_timer_init (&mytimer, timeout_cb, 0., 10.); /* note, only repeat used */
		2271	ev_timer_again (&mytimer); /* start timer */
		2272	ev_run (loop, 0);
		2273
		2274	// and in some piece of code that gets executed on any "activity":
		2275	// reset the timeout to start ticking again at 10 seconds
		2276	ev_timer_again (&mytimer);
		2277
		2278
526	=head2 C<ev_periodic> - to cron or not to cron	2279	=head2 C<ev_periodic> - to cron or not to cron?
527		2280
528	Periodic watchers are also timers of a kind, but they are very versatile	2281	Periodic watchers are also timers of a kind, but they are very versatile
529	(and unfortunately a bit complex).	2282	(and unfortunately a bit complex).
530		2283
531	Unlike C<ev_timer>'s, they are not based on real time (or relative time)	2284	Unlike C<ev_timer>, periodic watchers are not based on real time (or
532	but on wallclock time (absolute time). You can tell a periodic watcher	2285	relative time, the physical time that passes) but on wall clock time
533	to trigger "at" some specific point in time. For example, if you tell a	2286	(absolute time, the thing you can read on your calendar or clock). The
534	periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now ()	2287	difference is that wall clock time can run faster or slower than real
535	+ 10.>) and then reset your system clock to the last year, then it will	2288	time, and time jumps are not uncommon (e.g. when you adjust your
536	take a year to trigger the event (unlike an C<ev_timer>, which would trigger	2289	wrist-watch).
537	roughly 10 seconds later and of course not if you reset your system time
538	again).
539		2290
540	They can also be used to implement vastly more complex timers, such as	2291	You can tell a periodic watcher to trigger after some specific point
		2292	in time: for example, if you tell a periodic watcher to trigger "in 10
		2293	seconds" (by specifying e.g. C<ev_now () + 10.>, that is, an absolute time
		2294	not a delay) and then reset your system clock to January of the previous
		2295	year, then it will take a year or more to trigger the event (unlike an
		2296	C<ev_timer>, which would still trigger roughly 10 seconds after starting
		2297	it, as it uses a relative timeout).
		2298
		2299	C<ev_periodic> watchers can also be used to implement vastly more complex
541	triggering an event on eahc midnight, local time.	2300	timers, such as triggering an event on each "midnight, local time", or
		2301	other complicated rules. This cannot easily be done with C<ev_timer>
		2302	watchers, as those cannot react to time jumps.
		2303
		2304	As with timers, the callback is guaranteed to be invoked only when the
		2305	point in time where it is supposed to trigger has passed. If multiple
		2306	timers become ready during the same loop iteration then the ones with
		2307	earlier time-out values are invoked before ones with later time-out values
		2308	(but this is no longer true when a callback calls C<ev_run> recursively).
		2309
		2310	=head3 Watcher-Specific Functions and Data Members
542		2311
543	=over 4	2312	=over 4
544		2313
545	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)	2314	=item ev_periodic_init (ev_periodic *, callback, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
546		2315
547	=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)	2316	=item ev_periodic_set (ev_periodic *, ev_tstamp offset, ev_tstamp interval, reschedule_cb)
548		2317
549	Lots of arguments, lets sort it out... There are basically three modes of	2318	Lots of arguments, let's sort it out... There are basically three modes of
550	operation, and we will explain them from simplest to complex:	2319	operation, and we will explain them from simplest to most complex:
551
552		2320
553	=over 4	2321	=over 4
554		2322
555	=item * absolute timer (interval = reschedule_cb = 0)	2323	=item * absolute timer (offset = absolute time, interval = 0, reschedule_cb = 0)
556		2324
557	In this configuration the watcher triggers an event at the wallclock time	2325	In this configuration the watcher triggers an event after the wall clock
558	C<at> and doesn't repeat. It will not adjust when a time jump occurs,	2326	time C<offset> has passed. It will not repeat and will not adjust when a
559	that is, if it is to be run at January 1st 2011 then it will run when the	2327	time jump occurs, that is, if it is to be run at January 1st 2011 then it
560	system time reaches or surpasses this time.	2328	will be stopped and invoked when the system clock reaches or surpasses
		2329	this point in time.
561		2330
562	=item * non-repeating interval timer (interval > 0, reschedule_cb = 0)	2331	=item * repeating interval timer (offset = offset within interval, interval > 0, reschedule_cb = 0)
563		2332
564	In this mode the watcher will always be scheduled to time out at the next	2333	In this mode the watcher will always be scheduled to time out at the next
565	C<at + N * interval> time (for some integer N) and then repeat, regardless	2334	C<offset + N * interval> time (for some integer N, which can also be
566	of any time jumps.	2335	negative) and then repeat, regardless of any time jumps. The C<offset>
		2336	argument is merely an offset into the C<interval> periods.
567		2337
568	This can be used to create timers that do not drift with respect to system	2338	This can be used to create timers that do not drift with respect to the
569	time:	2339	system clock, for example, here is an C<ev_periodic> that triggers each
		2340	hour, on the hour (with respect to UTC):
570		2341
571	ev_periodic_set (&periodic, 0., 3600., 0);	2342	ev_periodic_set (&periodic, 0., 3600., 0);
572		2343
573	This doesn't mean there will always be 3600 seconds in between triggers,	2344	This doesn't mean there will always be 3600 seconds in between triggers,
574	but only that the the callback will be called when the system time shows a	2345	but only that the callback will be called when the system time shows a
575	full hour (UTC), or more correctly, when the system time is evenly divisible	2346	full hour (UTC), or more correctly, when the system time is evenly divisible
576	by 3600.	2347	by 3600.
577		2348
578	Another way to think about it (for the mathematically inclined) is that	2349	Another way to think about it (for the mathematically inclined) is that
579	C<ev_periodic> will try to run the callback in this mode at the next possible	2350	C<ev_periodic> will try to run the callback in this mode at the next possible
580	time where C<time = at (mod interval)>, regardless of any time jumps.	2351	time where C<time = offset (mod interval)>, regardless of any time jumps.
581		2352
		2353	The C<interval> I<MUST> be positive, and for numerical stability, the
		2354	interval value should be higher than C<1/8192> (which is around 100
		2355	microseconds) and C<offset> should be higher than C<0> and should have
		2356	at most a similar magnitude as the current time (say, within a factor of
		2357	ten). Typical values for offset are, in fact, C<0> or something between
		2358	C<0> and C<interval>, which is also the recommended range.
		2359
		2360	Note also that there is an upper limit to how often a timer can fire (CPU
		2361	speed for example), so if C<interval> is very small then timing stability
		2362	will of course deteriorate. Libev itself tries to be exact to be about one
		2363	millisecond (if the OS supports it and the machine is fast enough).
		2364
582	=item * manual reschedule mode (reschedule_cb = callback)	2365	=item * manual reschedule mode (offset ignored, interval ignored, reschedule_cb = callback)
583		2366
584	In this mode the values for C<interval> and C<at> are both being	2367	In this mode the values for C<interval> and C<offset> are both being
585	ignored. Instead, each time the periodic watcher gets scheduled, the	2368	ignored. Instead, each time the periodic watcher gets scheduled, the
586	reschedule callback will be called with the watcher as first, and the	2369	reschedule callback will be called with the watcher as first, and the
587	current time as second argument.	2370	current time as second argument.
588		2371
589	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher,	2372	NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, ever,
590	ever, or make any event loop modifications>. If you need to stop it,	2373	or make ANY other event loop modifications whatsoever, unless explicitly
591	return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by	2374	allowed by documentation here>.
592	starting a prepare watcher).
593		2375
		2376	If you need to stop it, return C<now + 1e30> (or so, fudge fudge) and stop
		2377	it afterwards (e.g. by starting an C<ev_prepare> watcher, which is the
		2378	only event loop modification you are allowed to do).
		2379
594	Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w,	2380	The callback prototype is C<ev_tstamp (*reschedule_cb)(ev_periodic
595	ev_tstamp now)>, e.g.:	2381	*w, ev_tstamp now)>, e.g.:
596		2382
		2383	static ev_tstamp
597	static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)	2384	my_rescheduler (ev_periodic *w, ev_tstamp now)
598	{	2385	{
599	return now + 60.;	2386	return now + 60.;
600	}	2387	}
601		2388
602	It must return the next time to trigger, based on the passed time value	2389	It must return the next time to trigger, based on the passed time value
603	(that is, the lowest time value larger than to the second argument). It	2390	(that is, the lowest time value larger than to the second argument). It
604	will usually be called just before the callback will be triggered, but	2391	will usually be called just before the callback will be triggered, but
605	might be called at other times, too.	2392	might be called at other times, too.
606		2393
607	NOTE: I<< This callback must always return a time that is later than the	2394	NOTE: I<< This callback must always return a time that is higher than or
608	passed C<now> value >>. Not even C<now> itself will do, it I<must> be larger.	2395	equal to the passed C<now> value >>.
609		2396
610	This can be used to create very complex timers, such as a timer that	2397	This can be used to create very complex timers, such as a timer that
611	triggers on each midnight, local time. To do this, you would calculate the	2398	triggers on "next midnight, local time". To do this, you would calculate
612	next midnight after C<now> and return the timestamp value for this. How	2399	the next midnight after C<now> and return the timestamp value for
613	you do this is, again, up to you (but it is not trivial, which is the main	2400	this. Here is a (completely untested, no error checking) example on how to
614	reason I omitted it as an example).	2401	do this:
		2402
		2403	#include <time.h>
		2404
		2405	static ev_tstamp
		2406	my_rescheduler (ev_periodic *w, ev_tstamp now)
		2407	{
		2408	time_t tnow = (time_t)now;
		2409	struct tm tm;
		2410	localtime_r (&tnow, &tm);
		2411
		2412	tm.tm_sec = tm.tm_min = tm.tm_hour = 0; // midnight current day
		2413	++tm.tm_mday; // midnight next day
		2414
		2415	return mktime (&tm);
		2416	}
		2417
		2418	Note: this code might run into trouble on days that have more then two
		2419	midnights (beginning and end).
615		2420
616	=back	2421	=back
617		2422
618	=item ev_periodic_again (loop, ev_periodic *)	2423	=item ev_periodic_again (loop, ev_periodic *)
619		2424
620	Simply stops and restarts the periodic watcher again. This is only useful	2425	Simply stops and restarts the periodic watcher again. This is only useful
621	when you changed some parameters or the reschedule callback would return	2426	when you changed some parameters or the reschedule callback would return
622	a different time than the last time it was called (e.g. in a crond like	2427	a different time than the last time it was called (e.g. in a crond like
623	program when the crontabs have changed).	2428	program when the crontabs have changed).
624		2429
		2430	=item ev_tstamp ev_periodic_at (ev_periodic *)
		2431
		2432	When active, returns the absolute time that the watcher is supposed
		2433	to trigger next. This is not the same as the C<offset> argument to
		2434	C<ev_periodic_set>, but indeed works even in interval and manual
		2435	rescheduling modes.
		2436
		2437	=item ev_tstamp offset [read-write]
		2438
		2439	When repeating, this contains the offset value, otherwise this is the
		2440	absolute point in time (the C<offset> value passed to C<ev_periodic_set>,
		2441	although libev might modify this value for better numerical stability).
		2442
		2443	Can be modified any time, but changes only take effect when the periodic
		2444	timer fires or C<ev_periodic_again> is being called.
		2445
		2446	=item ev_tstamp interval [read-write]
		2447
		2448	The current interval value. Can be modified any time, but changes only
		2449	take effect when the periodic timer fires or C<ev_periodic_again> is being
		2450	called.
		2451
		2452	=item ev_tstamp (*reschedule_cb)(ev_periodic *w, ev_tstamp now) [read-write]
		2453
		2454	The current reschedule callback, or C<0>, if this functionality is
		2455	switched off. Can be changed any time, but changes only take effect when
		2456	the periodic timer fires or C<ev_periodic_again> is being called.
		2457
625	=back	2458	=back
626		2459
		2460	=head3 Examples
		2461
		2462	Example: Call a callback every hour, or, more precisely, whenever the
		2463	system time is divisible by 3600. The callback invocation times have
		2464	potentially a lot of jitter, but good long-term stability.
		2465
		2466	static void
		2467	clock_cb (struct ev_loop *loop, ev_periodic *w, int revents)
		2468	{
		2469	... its now a full hour (UTC, or TAI or whatever your clock follows)
		2470	}
		2471
		2472	ev_periodic hourly_tick;
		2473	ev_periodic_init (&hourly_tick, clock_cb, 0., 3600., 0);
		2474	ev_periodic_start (loop, &hourly_tick);
		2475
		2476	Example: The same as above, but use a reschedule callback to do it:
		2477
		2478	#include <math.h>
		2479
		2480	static ev_tstamp
		2481	my_scheduler_cb (ev_periodic *w, ev_tstamp now)
		2482	{
		2483	return now + (3600. - fmod (now, 3600.));
		2484	}
		2485
		2486	ev_periodic_init (&hourly_tick, clock_cb, 0., 0., my_scheduler_cb);
		2487
		2488	Example: Call a callback every hour, starting now:
		2489
		2490	ev_periodic hourly_tick;
		2491	ev_periodic_init (&hourly_tick, clock_cb,
		2492	fmod (ev_now (loop), 3600.), 3600., 0);
		2493	ev_periodic_start (loop, &hourly_tick);
		2494
		2495
627	=head2 C<ev_signal> - signal me when a signal gets signalled	2496	=head2 C<ev_signal> - signal me when a signal gets signalled!
628		2497
629	Signal watchers will trigger an event when the process receives a specific	2498	Signal watchers will trigger an event when the process receives a specific
630	signal one or more times. Even though signals are very asynchronous, libev	2499	signal one or more times. Even though signals are very asynchronous, libev
631	will try it's best to deliver signals synchronously, i.e. as part of the	2500	will try its best to deliver signals synchronously, i.e. as part of the
632	normal event processing, like any other event.	2501	normal event processing, like any other event.
633		2502
		2503	If you want signals to be delivered truly asynchronously, just use
		2504	C<sigaction> as you would do without libev and forget about sharing
		2505	the signal. You can even use C<ev_async> from a signal handler to
		2506	synchronously wake up an event loop.
		2507
634	You can configure as many watchers as you like per signal. Only when the	2508	You can configure as many watchers as you like for the same signal, but
635	first watcher gets started will libev actually register a signal watcher	2509	only within the same loop, i.e. you can watch for C<SIGINT> in your
636	with the kernel (thus it coexists with your own signal handlers as long	2510	default loop and for C<SIGIO> in another loop, but you cannot watch for
637	as you don't register any with libev). Similarly, when the last signal	2511	C<SIGINT> in both the default loop and another loop at the same time. At
638	watcher for a signal is stopped libev will reset the signal handler to	2512	the moment, C<SIGCHLD> is permanently tied to the default loop.
639	SIG_DFL (regardless of what it was set to before).	2513
		2514	Only after the first watcher for a signal is started will libev actually
		2515	register something with the kernel. It thus coexists with your own signal
		2516	handlers as long as you don't register any with libev for the same signal.
		2517
		2518	If possible and supported, libev will install its handlers with
		2519	C<SA_RESTART> (or equivalent) behaviour enabled, so system calls should
		2520	not be unduly interrupted. If you have a problem with system calls getting
		2521	interrupted by signals you can block all signals in an C<ev_check> watcher
		2522	and unblock them in an C<ev_prepare> watcher.
		2523
		2524	=head3 The special problem of inheritance over fork/execve/pthread_create
		2525
		2526	Both the signal mask (C<sigprocmask>) and the signal disposition
		2527	(C<sigaction>) are unspecified after starting a signal watcher (and after
		2528	stopping it again), that is, libev might or might not block the signal,
		2529	and might or might not set or restore the installed signal handler (but
		2530	see C<EVFLAG_NOSIGMASK>).
		2531
		2532	While this does not matter for the signal disposition (libev never
		2533	sets signals to C<SIG_IGN>, so handlers will be reset to C<SIG_DFL> on
		2534	C<execve>), this matters for the signal mask: many programs do not expect
		2535	certain signals to be blocked.
		2536
		2537	This means that before calling C<exec> (from the child) you should reset
		2538	the signal mask to whatever "default" you expect (all clear is a good
		2539	choice usually).
		2540
		2541	The simplest way to ensure that the signal mask is reset in the child is
		2542	to install a fork handler with C<pthread_atfork> that resets it. That will
		2543	catch fork calls done by libraries (such as the libc) as well.
		2544
		2545	In current versions of libev, the signal will not be blocked indefinitely
		2546	unless you use the C<signalfd> API (C<EV_SIGNALFD>). While this reduces
		2547	the window of opportunity for problems, it will not go away, as libev
		2548	I<has> to modify the signal mask, at least temporarily.
		2549
		2550	So I can't stress this enough: I<If you do not reset your signal mask when
		2551	you expect it to be empty, you have a race condition in your code>. This
		2552	is not a libev-specific thing, this is true for most event libraries.
		2553
		2554	=head3 The special problem of threads signal handling
		2555
		2556	POSIX threads has problematic signal handling semantics, specifically,
		2557	a lot of functionality (sigfd, sigwait etc.) only really works if all
		2558	threads in a process block signals, which is hard to achieve.
		2559
		2560	When you want to use sigwait (or mix libev signal handling with your own
		2561	for the same signals), you can tackle this problem by globally blocking
		2562	all signals before creating any threads (or creating them with a fully set
		2563	sigprocmask) and also specifying the C<EVFLAG_NOSIGMASK> when creating
		2564	loops. Then designate one thread as "signal receiver thread" which handles
		2565	these signals. You can pass on any signals that libev might be interested
		2566	in by calling C<ev_feed_signal>.
		2567
		2568	=head3 Watcher-Specific Functions and Data Members
640		2569
641	=over 4	2570	=over 4
642		2571
643	=item ev_signal_init (ev_signal *, callback, int signum)	2572	=item ev_signal_init (ev_signal *, callback, int signum)
644		2573
645	=item ev_signal_set (ev_signal *, int signum)	2574	=item ev_signal_set (ev_signal *, int signum)
646		2575
647	Configures the watcher to trigger on the given signal number (usually one	2576	Configures the watcher to trigger on the given signal number (usually one
648	of the C<SIGxxx> constants).	2577	of the C<SIGxxx> constants).
649		2578
		2579	=item int signum [read-only]
		2580
		2581	The signal the watcher watches out for.
		2582
650	=back	2583	=back
651		2584
		2585	=head3 Examples
		2586
		2587	Example: Try to exit cleanly on SIGINT.
		2588
		2589	static void
		2590	sigint_cb (struct ev_loop *loop, ev_signal *w, int revents)
		2591	{
		2592	ev_break (loop, EVBREAK_ALL);
		2593	}
		2594
		2595	ev_signal signal_watcher;
		2596	ev_signal_init (&signal_watcher, sigint_cb, SIGINT);
		2597	ev_signal_start (loop, &signal_watcher);
		2598
		2599
652	=head2 C<ev_child> - wait for pid status changes	2600	=head2 C<ev_child> - watch out for process status changes
653		2601
654	Child watchers trigger when your process receives a SIGCHLD in response to	2602	Child watchers trigger when your process receives a SIGCHLD in response to
655	some child status changes (most typically when a child of yours dies).	2603	some child status changes (most typically when a child of yours dies or
		2604	exits). It is permissible to install a child watcher I<after> the child
		2605	has been forked (which implies it might have already exited), as long
		2606	as the event loop isn't entered (or is continued from a watcher), i.e.,
		2607	forking and then immediately registering a watcher for the child is fine,
		2608	but forking and registering a watcher a few event loop iterations later or
		2609	in the next callback invocation is not.
		2610
		2611	Only the default event loop is capable of handling signals, and therefore
		2612	you can only register child watchers in the default event loop.
		2613
		2614	Due to some design glitches inside libev, child watchers will always be
		2615	handled at maximum priority (their priority is set to C<EV_MAXPRI> by
		2616	libev)
		2617
		2618	=head3 Process Interaction
		2619
		2620	Libev grabs C<SIGCHLD> as soon as the default event loop is
		2621	initialised. This is necessary to guarantee proper behaviour even if the
		2622	first child watcher is started after the child exits. The occurrence
		2623	of C<SIGCHLD> is recorded asynchronously, but child reaping is done
		2624	synchronously as part of the event loop processing. Libev always reaps all
		2625	children, even ones not watched.
		2626
		2627	=head3 Overriding the Built-In Processing
		2628
		2629	Libev offers no special support for overriding the built-in child
		2630	processing, but if your application collides with libev's default child
		2631	handler, you can override it easily by installing your own handler for
		2632	C<SIGCHLD> after initialising the default loop, and making sure the
		2633	default loop never gets destroyed. You are encouraged, however, to use an
		2634	event-based approach to child reaping and thus use libev's support for
		2635	that, so other libev users can use C<ev_child> watchers freely.
		2636
		2637	=head3 Stopping the Child Watcher
		2638
		2639	Currently, the child watcher never gets stopped, even when the
		2640	child terminates, so normally one needs to stop the watcher in the
		2641	callback. Future versions of libev might stop the watcher automatically
		2642	when a child exit is detected (calling C<ev_child_stop> twice is not a
		2643	problem).
		2644
		2645	=head3 Watcher-Specific Functions and Data Members
656		2646
657	=over 4	2647	=over 4
658		2648
659	=item ev_child_init (ev_child *, callback, int pid)	2649	=item ev_child_init (ev_child *, callback, int pid, int trace)
660		2650
661	=item ev_child_set (ev_child *, int pid)	2651	=item ev_child_set (ev_child *, int pid, int trace)
662		2652
663	Configures the watcher to wait for status changes of process C<pid> (or	2653	Configures the watcher to wait for status changes of process C<pid> (or
664	I<any> process if C<pid> is specified as C<0>). The callback can look	2654	I<any> process if C<pid> is specified as C<0>). The callback can look
665	at the C<rstatus> member of the C<ev_child> watcher structure to see	2655	at the C<rstatus> member of the C<ev_child> watcher structure to see
666	the status word (use the macros from C<sys/wait.h> and see your systems	2656	the status word (use the macros from C<sys/wait.h> and see your systems
667	C<waitpid> documentation). The C<rpid> member contains the pid of the	2657	C<waitpid> documentation). The C<rpid> member contains the pid of the
668	process causing the status change.	2658	process causing the status change. C<trace> must be either C<0> (only
		2659	activate the watcher when the process terminates) or C<1> (additionally
		2660	activate the watcher when the process is stopped or continued).
		2661
		2662	=item int pid [read-only]
		2663
		2664	The process id this watcher watches out for, or C<0>, meaning any process id.
		2665
		2666	=item int rpid [read-write]
		2667
		2668	The process id that detected a status change.
		2669
		2670	=item int rstatus [read-write]
		2671
		2672	The process exit/trace status caused by C<rpid> (see your systems
		2673	C<waitpid> and C<sys/wait.h> documentation for details).
669		2674
670	=back	2675	=back
671		2676
		2677	=head3 Examples
		2678
		2679	Example: C<fork()> a new process and install a child handler to wait for
		2680	its completion.
		2681
		2682	ev_child cw;
		2683
		2684	static void
		2685	child_cb (EV_P_ ev_child *w, int revents)
		2686	{
		2687	ev_child_stop (EV_A_ w);
		2688	printf ("process %d exited with status %x\n", w->rpid, w->rstatus);
		2689	}
		2690
		2691	pid_t pid = fork ();
		2692
		2693	if (pid < 0)
		2694	// error
		2695	else if (pid == 0)
		2696	{
		2697	// the forked child executes here
		2698	exit (1);
		2699	}
		2700	else
		2701	{
		2702	ev_child_init (&cw, child_cb, pid, 0);
		2703	ev_child_start (EV_DEFAULT_ &cw);
		2704	}
		2705
		2706
		2707	=head2 C<ev_stat> - did the file attributes just change?
		2708
		2709	This watches a file system path for attribute changes. That is, it calls
		2710	C<stat> on that path in regular intervals (or when the OS says it changed)
		2711	and sees if it changed compared to the last time, invoking the callback
		2712	if it did. Starting the watcher C<stat>'s the file, so only changes that
		2713	happen after the watcher has been started will be reported.
		2714
		2715	The path does not need to exist: changing from "path exists" to "path does
		2716	not exist" is a status change like any other. The condition "path does not
		2717	exist" (or more correctly "path cannot be stat'ed") is signified by the
		2718	C<st_nlink> field being zero (which is otherwise always forced to be at
		2719	least one) and all the other fields of the stat buffer having unspecified
		2720	contents.
		2721
		2722	The path I<must not> end in a slash or contain special components such as
		2723	C<.> or C<..>. The path I<should> be absolute: If it is relative and
		2724	your working directory changes, then the behaviour is undefined.
		2725
		2726	Since there is no portable change notification interface available, the
		2727	portable implementation simply calls C<stat(2)> regularly on the path
		2728	to see if it changed somehow. You can specify a recommended polling
		2729	interval for this case. If you specify a polling interval of C<0> (highly
		2730	recommended!) then a I<suitable, unspecified default> value will be used
		2731	(which you can expect to be around five seconds, although this might
		2732	change dynamically). Libev will also impose a minimum interval which is
		2733	currently around C<0.1>, but that's usually overkill.
		2734
		2735	This watcher type is not meant for massive numbers of stat watchers,
		2736	as even with OS-supported change notifications, this can be
		2737	resource-intensive.
		2738
		2739	At the time of this writing, the only OS-specific interface implemented
		2740	is the Linux inotify interface (implementing kqueue support is left as an
		2741	exercise for the reader. Note, however, that the author sees no way of
		2742	implementing C<ev_stat> semantics with kqueue, except as a hint).
		2743
		2744	=head3 ABI Issues (Largefile Support)
		2745
		2746	Libev by default (unless the user overrides this) uses the default
		2747	compilation environment, which means that on systems with large file
		2748	support disabled by default, you get the 32 bit version of the stat
		2749	structure. When using the library from programs that change the ABI to
		2750	use 64 bit file offsets the programs will fail. In that case you have to
		2751	compile libev with the same flags to get binary compatibility. This is
		2752	obviously the case with any flags that change the ABI, but the problem is
		2753	most noticeably displayed with ev_stat and large file support.
		2754
		2755	The solution for this is to lobby your distribution maker to make large
		2756	file interfaces available by default (as e.g. FreeBSD does) and not
		2757	optional. Libev cannot simply switch on large file support because it has
		2758	to exchange stat structures with application programs compiled using the
		2759	default compilation environment.
		2760
		2761	=head3 Inotify and Kqueue
		2762
		2763	When C<inotify (7)> support has been compiled into libev and present at
		2764	runtime, it will be used to speed up change detection where possible. The
		2765	inotify descriptor will be created lazily when the first C<ev_stat>
		2766	watcher is being started.
		2767
		2768	Inotify presence does not change the semantics of C<ev_stat> watchers
		2769	except that changes might be detected earlier, and in some cases, to avoid
		2770	making regular C<stat> calls. Even in the presence of inotify support
		2771	there are many cases where libev has to resort to regular C<stat> polling,
		2772	but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too
		2773	many bugs), the path exists (i.e. stat succeeds), and the path resides on
		2774	a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and
		2775	xfs are fully working) libev usually gets away without polling.
		2776
		2777	There is no support for kqueue, as apparently it cannot be used to
		2778	implement this functionality, due to the requirement of having a file
		2779	descriptor open on the object at all times, and detecting renames, unlinks
		2780	etc. is difficult.
		2781
		2782	=head3 C<stat ()> is a synchronous operation
		2783
		2784	Libev doesn't normally do any kind of I/O itself, and so is not blocking
		2785	the process. The exception are C<ev_stat> watchers - those call C<stat
		2786	()>, which is a synchronous operation.
		2787
		2788	For local paths, this usually doesn't matter: unless the system is very
		2789	busy or the intervals between stat's are large, a stat call will be fast,
		2790	as the path data is usually in memory already (except when starting the
		2791	watcher).
		2792
		2793	For networked file systems, calling C<stat ()> can block an indefinite
		2794	time due to network issues, and even under good conditions, a stat call
		2795	often takes multiple milliseconds.
		2796
		2797	Therefore, it is best to avoid using C<ev_stat> watchers on networked
		2798	paths, although this is fully supported by libev.
		2799
		2800	=head3 The special problem of stat time resolution
		2801
		2802	The C<stat ()> system call only supports full-second resolution portably,
		2803	and even on systems where the resolution is higher, most file systems
		2804	still only support whole seconds.
		2805
		2806	That means that, if the time is the only thing that changes, you can
		2807	easily miss updates: on the first update, C<ev_stat> detects a change and
		2808	calls your callback, which does something. When there is another update
		2809	within the same second, C<ev_stat> will be unable to detect unless the
		2810	stat data does change in other ways (e.g. file size).
		2811
		2812	The solution to this is to delay acting on a change for slightly more
		2813	than a second (or till slightly after the next full second boundary), using
		2814	a roughly one-second-delay C<ev_timer> (e.g. C<ev_timer_set (w, 0., 1.02);
		2815	ev_timer_again (loop, w)>).
		2816
		2817	The C<.02> offset is added to work around small timing inconsistencies
		2818	of some operating systems (where the second counter of the current time
		2819	might be be delayed. One such system is the Linux kernel, where a call to
		2820	C<gettimeofday> might return a timestamp with a full second later than
		2821	a subsequent C<time> call - if the equivalent of C<time ()> is used to
		2822	update file times then there will be a small window where the kernel uses
		2823	the previous second to update file times but libev might already execute
		2824	the timer callback).
		2825
		2826	=head3 Watcher-Specific Functions and Data Members
		2827
		2828	=over 4
		2829
		2830	=item ev_stat_init (ev_stat *, callback, const char *path, ev_tstamp interval)
		2831
		2832	=item ev_stat_set (ev_stat *, const char *path, ev_tstamp interval)
		2833
		2834	Configures the watcher to wait for status changes of the given
		2835	C<path>. The C<interval> is a hint on how quickly a change is expected to
		2836	be detected and should normally be specified as C<0> to let libev choose
		2837	a suitable value. The memory pointed to by C<path> must point to the same
		2838	path for as long as the watcher is active.
		2839
		2840	The callback will receive an C<EV_STAT> event when a change was detected,
		2841	relative to the attributes at the time the watcher was started (or the
		2842	last change was detected).
		2843
		2844	=item ev_stat_stat (loop, ev_stat *)
		2845
		2846	Updates the stat buffer immediately with new values. If you change the
		2847	watched path in your callback, you could call this function to avoid
		2848	detecting this change (while introducing a race condition if you are not
		2849	the only one changing the path). Can also be useful simply to find out the
		2850	new values.
		2851
		2852	=item ev_statdata attr [read-only]
		2853
		2854	The most-recently detected attributes of the file. Although the type is
		2855	C<ev_statdata>, this is usually the (or one of the) C<struct stat> types
		2856	suitable for your system, but you can only rely on the POSIX-standardised
		2857	members to be present. If the C<st_nlink> member is C<0>, then there was
		2858	some error while C<stat>ing the file.
		2859
		2860	=item ev_statdata prev [read-only]
		2861
		2862	The previous attributes of the file. The callback gets invoked whenever
		2863	C<prev> != C<attr>, or, more precisely, one or more of these members
		2864	differ: C<st_dev>, C<st_ino>, C<st_mode>, C<st_nlink>, C<st_uid>,
		2865	C<st_gid>, C<st_rdev>, C<st_size>, C<st_atime>, C<st_mtime>, C<st_ctime>.
		2866
		2867	=item ev_tstamp interval [read-only]
		2868
		2869	The specified interval.
		2870
		2871	=item const char *path [read-only]
		2872
		2873	The file system path that is being watched.
		2874
		2875	=back
		2876
		2877	=head3 Examples
		2878
		2879	Example: Watch C</etc/passwd> for attribute changes.
		2880
		2881	static void
		2882	passwd_cb (struct ev_loop *loop, ev_stat *w, int revents)
		2883	{
		2884	/* /etc/passwd changed in some way */
		2885	if (w->attr.st_nlink)
		2886	{
		2887	printf ("passwd current size %ld\n", (long)w->attr.st_size);
		2888	printf ("passwd current atime %ld\n", (long)w->attr.st_mtime);
		2889	printf ("passwd current mtime %ld\n", (long)w->attr.st_mtime);
		2890	}
		2891	else
		2892	/* you shalt not abuse printf for puts */
		2893	puts ("wow, /etc/passwd is not there, expect problems. "
		2894	"if this is windows, they already arrived\n");
		2895	}
		2896
		2897	...
		2898	ev_stat passwd;
		2899
		2900	ev_stat_init (&passwd, passwd_cb, "/etc/passwd", 0.);
		2901	ev_stat_start (loop, &passwd);
		2902
		2903	Example: Like above, but additionally use a one-second delay so we do not
		2904	miss updates (however, frequent updates will delay processing, too, so
		2905	one might do the work both on C<ev_stat> callback invocation I<and> on
		2906	C<ev_timer> callback invocation).
		2907
		2908	static ev_stat passwd;
		2909	static ev_timer timer;
		2910
		2911	static void
		2912	timer_cb (EV_P_ ev_timer *w, int revents)
		2913	{
		2914	ev_timer_stop (EV_A_ w);
		2915
		2916	/* now it's one second after the most recent passwd change */
		2917	}
		2918
		2919	static void
		2920	stat_cb (EV_P_ ev_stat *w, int revents)
		2921	{
		2922	/* reset the one-second timer */
		2923	ev_timer_again (EV_A_ &timer);
		2924	}
		2925
		2926	...
		2927	ev_stat_init (&passwd, stat_cb, "/etc/passwd", 0.);
		2928	ev_stat_start (loop, &passwd);
		2929	ev_timer_init (&timer, timer_cb, 0., 1.02);
		2930
		2931
672	=head2 C<ev_idle> - when you've got nothing better to do	2932	=head2 C<ev_idle> - when you've got nothing better to do...
673		2933
674	Idle watchers trigger events when there are no other events are pending	2934	Idle watchers trigger events when no other events of the same or higher
675	(prepare, check and other idle watchers do not count). That is, as long	2935	priority are pending (prepare, check and other idle watchers do not count
676	as your process is busy handling sockets or timeouts (or even signals,	2936	as receiving "events").
677	imagine) it will not be triggered. But when your process is idle all idle	2937
678	watchers are being called again and again, once per event loop iteration -	2938	That is, as long as your process is busy handling sockets or timeouts
		2939	(or even signals, imagine) of the same or higher priority it will not be
		2940	triggered. But when your process is idle (or only lower-priority watchers
		2941	are pending), the idle watchers are being called once per event loop
679	until stopped, that is, or your process receives more events and becomes	2942	iteration - until stopped, that is, or your process receives more events
680	busy.	2943	and becomes busy again with higher priority stuff.
681		2944
682	The most noteworthy effect is that as long as any idle watchers are	2945	The most noteworthy effect is that as long as any idle watchers are
683	active, the process will not block when waiting for new events.	2946	active, the process will not block when waiting for new events.
684		2947
685	Apart from keeping your process non-blocking (which is a useful	2948	Apart from keeping your process non-blocking (which is a useful
686	effect on its own sometimes), idle watchers are a good place to do	2949	effect on its own sometimes), idle watchers are a good place to do
687	"pseudo-background processing", or delay processing stuff to after the	2950	"pseudo-background processing", or delay processing stuff to after the
688	event loop has handled all outstanding events.	2951	event loop has handled all outstanding events.
689		2952
		2953	=head3 Abusing an C<ev_idle> watcher for its side-effect
		2954
		2955	As long as there is at least one active idle watcher, libev will never
		2956	sleep unnecessarily. Or in other words, it will loop as fast as possible.
		2957	For this to work, the idle watcher doesn't need to be invoked at all - the
		2958	lowest priority will do.
		2959
		2960	This mode of operation can be useful together with an C<ev_check> watcher,
		2961	to do something on each event loop iteration - for example to balance load
		2962	between different connections.
		2963
		2964	See L</Abusing an ev_check watcher for its side-effect> for a longer
		2965	example.
		2966
		2967	=head3 Watcher-Specific Functions and Data Members
		2968
690	=over 4	2969	=over 4
691		2970
692	=item ev_idle_init (ev_signal *, callback)	2971	=item ev_idle_init (ev_idle *, callback)
693		2972
694	Initialises and configures the idle watcher - it has no parameters of any	2973	Initialises and configures the idle watcher - it has no parameters of any
695	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,	2974	kind. There is a C<ev_idle_set> macro, but using it is utterly pointless,
696	believe me.	2975	believe me.
697		2976
698	=back	2977	=back
699		2978
		2979	=head3 Examples
		2980
		2981	Example: Dynamically allocate an C<ev_idle> watcher, start it, and in the
		2982	callback, free it. Also, use no error checking, as usual.
		2983
		2984	static void
		2985	idle_cb (struct ev_loop *loop, ev_idle *w, int revents)
		2986	{
		2987	// stop the watcher
		2988	ev_idle_stop (loop, w);
		2989
		2990	// now we can free it
		2991	free (w);
		2992
		2993	// now do something you wanted to do when the program has
		2994	// no longer anything immediate to do.
		2995	}
		2996
		2997	ev_idle *idle_watcher = malloc (sizeof (ev_idle));
		2998	ev_idle_init (idle_watcher, idle_cb);
		2999	ev_idle_start (loop, idle_watcher);
		3000
		3001
700	=head2 C<ev_prepare> and C<ev_check> - customise your event loop	3002	=head2 C<ev_prepare> and C<ev_check> - customise your event loop!
701		3003
702	Prepare and check watchers are usually (but not always) used in tandem:	3004	Prepare and check watchers are often (but not always) used in pairs:
703	prepare watchers get invoked before the process blocks and check watchers	3005	prepare watchers get invoked before the process blocks and check watchers
704	afterwards.	3006	afterwards.
705		3007
		3008	You I<must not> call C<ev_run> (or similar functions that enter the
		3009	current event loop) or C<ev_loop_fork> from either C<ev_prepare> or
		3010	C<ev_check> watchers. Other loops than the current one are fine,
		3011	however. The rationale behind this is that you do not need to check
		3012	for recursion in those watchers, i.e. the sequence will always be
		3013	C<ev_prepare>, blocking, C<ev_check> so if you have one watcher of each
		3014	kind they will always be called in pairs bracketing the blocking call.
		3015
706	Their main purpose is to integrate other event mechanisms into libev. This	3016	Their main purpose is to integrate other event mechanisms into libev and
707	could be used, for example, to track variable changes, implement your own	3017	their use is somewhat advanced. They could be used, for example, to track
708	watchers, integrate net-snmp or a coroutine library and lots more.	3018	variable changes, implement your own watchers, integrate net-snmp or a
		3019	coroutine library and lots more. They are also occasionally useful if
		3020	you cache some data and want to flush it before blocking (for example,
		3021	in X programs you might want to do an C<XFlush ()> in an C<ev_prepare>
		3022	watcher).
709		3023
710	This is done by examining in each prepare call which file descriptors need	3024	This is done by examining in each prepare call which file descriptors
711	to be watched by the other library, registering C<ev_io> watchers for	3025	need to be watched by the other library, registering C<ev_io> watchers
712	them and starting an C<ev_timer> watcher for any timeouts (many libraries	3026	for them and starting an C<ev_timer> watcher for any timeouts (many
713	provide just this functionality). Then, in the check watcher you check for	3027	libraries provide exactly this functionality). Then, in the check watcher,
714	any events that occured (by checking the pending status of all watchers	3028	you check for any events that occurred (by checking the pending status
715	and stopping them) and call back into the library. The I/O and timer	3029	of all watchers and stopping them) and call back into the library. The
716	callbacks will never actually be called (but must be valid nevertheless,	3030	I/O and timer callbacks will never actually be called (but must be valid
717	because you never know, you know?).	3031	nevertheless, because you never know, you know?).
718		3032
719	As another example, the Perl Coro module uses these hooks to integrate	3033	As another example, the Perl Coro module uses these hooks to integrate
720	coroutines into libev programs, by yielding to other active coroutines	3034	coroutines into libev programs, by yielding to other active coroutines
721	during each prepare and only letting the process block if no coroutines	3035	during each prepare and only letting the process block if no coroutines
722	are ready to run (it's actually more complicated: it only runs coroutines	3036	are ready to run (it's actually more complicated: it only runs coroutines
723	with priority higher than or equal to the event loop and one coroutine	3037	with priority higher than or equal to the event loop and one coroutine
724	of lower priority, but only once, using idle watchers to keep the event	3038	of lower priority, but only once, using idle watchers to keep the event
725	loop from blocking if lower-priority coroutines are active, thus mapping	3039	loop from blocking if lower-priority coroutines are active, thus mapping
726	low-priority coroutines to idle/background tasks).	3040	low-priority coroutines to idle/background tasks).
727		3041
		3042	When used for this purpose, it is recommended to give C<ev_check> watchers
		3043	highest (C<EV_MAXPRI>) priority, to ensure that they are being run before
		3044	any other watchers after the poll (this doesn't matter for C<ev_prepare>
		3045	watchers).
		3046
		3047	Also, C<ev_check> watchers (and C<ev_prepare> watchers, too) should not
		3048	activate ("feed") events into libev. While libev fully supports this, they
		3049	might get executed before other C<ev_check> watchers did their job. As
		3050	C<ev_check> watchers are often used to embed other (non-libev) event
		3051	loops those other event loops might be in an unusable state until their
		3052	C<ev_check> watcher ran (always remind yourself to coexist peacefully with
		3053	others).
		3054
		3055	=head3 Abusing an C<ev_check> watcher for its side-effect
		3056
		3057	C<ev_check> (and less often also C<ev_prepare>) watchers can also be
		3058	useful because they are called once per event loop iteration. For
		3059	example, if you want to handle a large number of connections fairly, you
		3060	normally only do a bit of work for each active connection, and if there
		3061	is more work to do, you wait for the next event loop iteration, so other
		3062	connections have a chance of making progress.
		3063
		3064	Using an C<ev_check> watcher is almost enough: it will be called on the
		3065	next event loop iteration. However, that isn't as soon as possible -
		3066	without external events, your C<ev_check> watcher will not be invoked.
		3067
		3068	This is where C<ev_idle> watchers come in handy - all you need is a
		3069	single global idle watcher that is active as long as you have one active
		3070	C<ev_check> watcher. The C<ev_idle> watcher makes sure the event loop
		3071	will not sleep, and the C<ev_check> watcher makes sure a callback gets
		3072	invoked. Neither watcher alone can do that.
		3073
		3074	=head3 Watcher-Specific Functions and Data Members
		3075
728	=over 4	3076	=over 4
729		3077
730	=item ev_prepare_init (ev_prepare *, callback)	3078	=item ev_prepare_init (ev_prepare *, callback)
731		3079
732	=item ev_check_init (ev_check *, callback)	3080	=item ev_check_init (ev_check *, callback)
733		3081
734	Initialises and configures the prepare or check watcher - they have no	3082	Initialises and configures the prepare or check watcher - they have no
735	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>	3083	parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set>
736	macros, but using them is utterly, utterly and completely pointless.	3084	macros, but using them is utterly, utterly, utterly and completely
		3085	pointless.
737		3086
738	=back	3087	=back
739		3088
		3089	=head3 Examples
		3090
		3091	There are a number of principal ways to embed other event loops or modules
		3092	into libev. Here are some ideas on how to include libadns into libev
		3093	(there is a Perl module named C<EV::ADNS> that does this, which you could
		3094	use as a working example. Another Perl module named C<EV::Glib> embeds a
		3095	Glib main context into libev, and finally, C<Glib::EV> embeds EV into the
		3096	Glib event loop).
		3097
		3098	Method 1: Add IO watchers and a timeout watcher in a prepare handler,
		3099	and in a check watcher, destroy them and call into libadns. What follows
		3100	is pseudo-code only of course. This requires you to either use a low
		3101	priority for the check watcher or use C<ev_clear_pending> explicitly, as
		3102	the callbacks for the IO/timeout watchers might not have been called yet.
		3103
		3104	static ev_io iow [nfd];
		3105	static ev_timer tw;
		3106
		3107	static void
		3108	io_cb (struct ev_loop *loop, ev_io *w, int revents)
		3109	{
		3110	}
		3111
		3112	// create io watchers for each fd and a timer before blocking
		3113	static void
		3114	adns_prepare_cb (struct ev_loop *loop, ev_prepare *w, int revents)
		3115	{
		3116	int timeout = 3600000;
		3117	struct pollfd fds [nfd];
		3118	// actual code will need to loop here and realloc etc.
		3119	adns_beforepoll (ads, fds, &nfd, &timeout, timeval_from (ev_time ()));
		3120
		3121	/* the callback is illegal, but won't be called as we stop during check */
		3122	ev_timer_init (&tw, 0, timeout * 1e-3, 0.);
		3123	ev_timer_start (loop, &tw);
		3124
		3125	// create one ev_io per pollfd
		3126	for (int i = 0; i < nfd; ++i)
		3127	{
		3128	ev_io_init (iow + i, io_cb, fds [i].fd,
		3129	((fds [i].events & POLLIN ? EV_READ : 0)
		3130	| (fds [i].events & POLLOUT ? EV_WRITE : 0)));
		3131
		3132	fds [i].revents = 0;
		3133	ev_io_start (loop, iow + i);
		3134	}
		3135	}
		3136
		3137	// stop all watchers after blocking
		3138	static void
		3139	adns_check_cb (struct ev_loop *loop, ev_check *w, int revents)
		3140	{
		3141	ev_timer_stop (loop, &tw);
		3142
		3143	for (int i = 0; i < nfd; ++i)
		3144	{
		3145	// set the relevant poll flags
		3146	// could also call adns_processreadable etc. here
		3147	struct pollfd *fd = fds + i;
		3148	int revents = ev_clear_pending (iow + i);
		3149	if (revents & EV_READ ) fd->revents |= fd->events & POLLIN;
		3150	if (revents & EV_WRITE) fd->revents |= fd->events & POLLOUT;
		3151
		3152	// now stop the watcher
		3153	ev_io_stop (loop, iow + i);
		3154	}
		3155
		3156	adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
		3157	}
		3158
		3159	Method 2: This would be just like method 1, but you run C<adns_afterpoll>
		3160	in the prepare watcher and would dispose of the check watcher.
		3161
		3162	Method 3: If the module to be embedded supports explicit event
		3163	notification (libadns does), you can also make use of the actual watcher
		3164	callbacks, and only destroy/create the watchers in the prepare watcher.
		3165
		3166	static void
		3167	timer_cb (EV_P_ ev_timer *w, int revents)
		3168	{
		3169	adns_state ads = (adns_state)w->data;
		3170	update_now (EV_A);
		3171
		3172	adns_processtimeouts (ads, &tv_now);
		3173	}
		3174
		3175	static void
		3176	io_cb (EV_P_ ev_io *w, int revents)
		3177	{
		3178	adns_state ads = (adns_state)w->data;
		3179	update_now (EV_A);
		3180
		3181	if (revents & EV_READ ) adns_processreadable (ads, w->fd, &tv_now);
		3182	if (revents & EV_WRITE) adns_processwriteable (ads, w->fd, &tv_now);
		3183	}
		3184
		3185	// do not ever call adns_afterpoll
		3186
		3187	Method 4: Do not use a prepare or check watcher because the module you
		3188	want to embed is not flexible enough to support it. Instead, you can
		3189	override their poll function. The drawback with this solution is that the
		3190	main loop is now no longer controllable by EV. The C<Glib::EV> module uses
		3191	this approach, effectively embedding EV as a client into the horrible
		3192	libglib event loop.
		3193
		3194	static gint
		3195	event_poll_func (GPollFD *fds, guint nfds, gint timeout)
		3196	{
		3197	int got_events = 0;
		3198
		3199	for (n = 0; n < nfds; ++n)
		3200	// create/start io watcher that sets the relevant bits in fds[n] and increment got_events
		3201
		3202	if (timeout >= 0)
		3203	// create/start timer
		3204
		3205	// poll
		3206	ev_run (EV_A_ 0);
		3207
		3208	// stop timer again
		3209	if (timeout >= 0)
		3210	ev_timer_stop (EV_A_ &to);
		3211
		3212	// stop io watchers again - their callbacks should have set
		3213	for (n = 0; n < nfds; ++n)
		3214	ev_io_stop (EV_A_ iow [n]);
		3215
		3216	return got_events;
		3217	}
		3218
		3219
		3220	=head2 C<ev_embed> - when one backend isn't enough...
		3221
		3222	This is a rather advanced watcher type that lets you embed one event loop
		3223	into another (currently only C<ev_io> events are supported in the embedded
		3224	loop, other types of watchers might be handled in a delayed or incorrect
		3225	fashion and must not be used).
		3226
		3227	There are primarily two reasons you would want that: work around bugs and
		3228	prioritise I/O.
		3229
		3230	As an example for a bug workaround, the kqueue backend might only support
		3231	sockets on some platform, so it is unusable as generic backend, but you
		3232	still want to make use of it because you have many sockets and it scales
		3233	so nicely. In this case, you would create a kqueue-based loop and embed
		3234	it into your default loop (which might use e.g. poll). Overall operation
		3235	will be a bit slower because first libev has to call C<poll> and then
		3236	C<kevent>, but at least you can use both mechanisms for what they are
		3237	best: C<kqueue> for scalable sockets and C<poll> if you want it to work :)
		3238
		3239	As for prioritising I/O: under rare circumstances you have the case where
		3240	some fds have to be watched and handled very quickly (with low latency),
		3241	and even priorities and idle watchers might have too much overhead. In
		3242	this case you would put all the high priority stuff in one loop and all
		3243	the rest in a second one, and embed the second one in the first.
		3244
		3245	As long as the watcher is active, the callback will be invoked every
		3246	time there might be events pending in the embedded loop. The callback
		3247	must then call C<ev_embed_sweep (mainloop, watcher)> to make a single
		3248	sweep and invoke their callbacks (the callback doesn't need to invoke the
		3249	C<ev_embed_sweep> function directly, it could also start an idle watcher
		3250	to give the embedded loop strictly lower priority for example).
		3251
		3252	You can also set the callback to C<0>, in which case the embed watcher
		3253	will automatically execute the embedded loop sweep whenever necessary.
		3254
		3255	Fork detection will be handled transparently while the C<ev_embed> watcher
		3256	is active, i.e., the embedded loop will automatically be forked when the
		3257	embedding loop forks. In other cases, the user is responsible for calling
		3258	C<ev_loop_fork> on the embedded loop.
		3259
		3260	Unfortunately, not all backends are embeddable: only the ones returned by
		3261	C<ev_embeddable_backends> are, which, unfortunately, does not include any
		3262	portable one.
		3263
		3264	So when you want to use this feature you will always have to be prepared
		3265	that you cannot get an embeddable loop. The recommended way to get around
		3266	this is to have a separate variables for your embeddable loop, try to
		3267	create it, and if that fails, use the normal loop for everything.
		3268
		3269	=head3 C<ev_embed> and fork
		3270
		3271	While the C<ev_embed> watcher is running, forks in the embedding loop will
		3272	automatically be applied to the embedded loop as well, so no special
		3273	fork handling is required in that case. When the watcher is not running,
		3274	however, it is still the task of the libev user to call C<ev_loop_fork ()>
		3275	as applicable.
		3276
		3277	=head3 Watcher-Specific Functions and Data Members
		3278
		3279	=over 4
		3280
		3281	=item ev_embed_init (ev_embed *, callback, struct ev_loop *embedded_loop)
		3282
		3283	=item ev_embed_set (ev_embed *, struct ev_loop *embedded_loop)
		3284
		3285	Configures the watcher to embed the given loop, which must be
		3286	embeddable. If the callback is C<0>, then C<ev_embed_sweep> will be
		3287	invoked automatically, otherwise it is the responsibility of the callback
		3288	to invoke it (it will continue to be called until the sweep has been done,
		3289	if you do not want that, you need to temporarily stop the embed watcher).
		3290
		3291	=item ev_embed_sweep (loop, ev_embed *)
		3292
		3293	Make a single, non-blocking sweep over the embedded loop. This works
		3294	similarly to C<ev_run (embedded_loop, EVRUN_NOWAIT)>, but in the most
		3295	appropriate way for embedded loops.
		3296
		3297	=item struct ev_loop *other [read-only]
		3298
		3299	The embedded event loop.
		3300
		3301	=back
		3302
		3303	=head3 Examples
		3304
		3305	Example: Try to get an embeddable event loop and embed it into the default
		3306	event loop. If that is not possible, use the default loop. The default
		3307	loop is stored in C<loop_hi>, while the embeddable loop is stored in
		3308	C<loop_lo> (which is C<loop_hi> in the case no embeddable loop can be
		3309	used).
		3310
		3311	struct ev_loop *loop_hi = ev_default_init (0);
		3312	struct ev_loop *loop_lo = 0;
		3313	ev_embed embed;
		3314
		3315	// see if there is a chance of getting one that works
		3316	// (remember that a flags value of 0 means autodetection)
		3317	loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
		3318	? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
		3319	: 0;
		3320
		3321	// if we got one, then embed it, otherwise default to loop_hi
		3322	if (loop_lo)
		3323	{
		3324	ev_embed_init (&embed, 0, loop_lo);
		3325	ev_embed_start (loop_hi, &embed);
		3326	}
		3327	else
		3328	loop_lo = loop_hi;
		3329
		3330	Example: Check if kqueue is available but not recommended and create
		3331	a kqueue backend for use with sockets (which usually work with any
		3332	kqueue implementation). Store the kqueue/socket-only event loop in
		3333	C<loop_socket>. (One might optionally use C<EVFLAG_NOENV>, too).
		3334
		3335	struct ev_loop *loop = ev_default_init (0);
		3336	struct ev_loop *loop_socket = 0;
		3337	ev_embed embed;
		3338
		3339	if (ev_supported_backends () & ~ev_recommended_backends () & EVBACKEND_KQUEUE)
		3340	if ((loop_socket = ev_loop_new (EVBACKEND_KQUEUE))
		3341	{
		3342	ev_embed_init (&embed, 0, loop_socket);
		3343	ev_embed_start (loop, &embed);
		3344	}
		3345
		3346	if (!loop_socket)
		3347	loop_socket = loop;
		3348
		3349	// now use loop_socket for all sockets, and loop for everything else
		3350
		3351
		3352	=head2 C<ev_fork> - the audacity to resume the event loop after a fork
		3353
		3354	Fork watchers are called when a C<fork ()> was detected (usually because
		3355	whoever is a good citizen cared to tell libev about it by calling
		3356	C<ev_loop_fork>). The invocation is done before the event loop blocks next
		3357	and before C<ev_check> watchers are being called, and only in the child
		3358	after the fork. If whoever good citizen calling C<ev_default_fork> cheats
		3359	and calls it in the wrong process, the fork handlers will be invoked, too,
		3360	of course.
		3361
		3362	=head3 The special problem of life after fork - how is it possible?
		3363
		3364	Most uses of C<fork ()> consist of forking, then some simple calls to set
		3365	up/change the process environment, followed by a call to C<exec()>. This
		3366	sequence should be handled by libev without any problems.
		3367
		3368	This changes when the application actually wants to do event handling
		3369	in the child, or both parent in child, in effect "continuing" after the
		3370	fork.
		3371
		3372	The default mode of operation (for libev, with application help to detect
		3373	forks) is to duplicate all the state in the child, as would be expected
		3374	when I<either> the parent I<or> the child process continues.
		3375
		3376	When both processes want to continue using libev, then this is usually the
		3377	wrong result. In that case, usually one process (typically the parent) is
		3378	supposed to continue with all watchers in place as before, while the other
		3379	process typically wants to start fresh, i.e. without any active watchers.
		3380
		3381	The cleanest and most efficient way to achieve that with libev is to
		3382	simply create a new event loop, which of course will be "empty", and
		3383	use that for new watchers. This has the advantage of not touching more
		3384	memory than necessary, and thus avoiding the copy-on-write, and the
		3385	disadvantage of having to use multiple event loops (which do not support
		3386	signal watchers).
		3387
		3388	When this is not possible, or you want to use the default loop for
		3389	other reasons, then in the process that wants to start "fresh", call
		3390	C<ev_loop_destroy (EV_DEFAULT)> followed by C<ev_default_loop (...)>.
		3391	Destroying the default loop will "orphan" (not stop) all registered
		3392	watchers, so you have to be careful not to execute code that modifies
		3393	those watchers. Note also that in that case, you have to re-register any
		3394	signal watchers.
		3395
		3396	=head3 Watcher-Specific Functions and Data Members
		3397
		3398	=over 4
		3399
		3400	=item ev_fork_init (ev_fork *, callback)
		3401
		3402	Initialises and configures the fork watcher - it has no parameters of any
		3403	kind. There is a C<ev_fork_set> macro, but using it is utterly pointless,
		3404	really.
		3405
		3406	=back
		3407
		3408
		3409	=head2 C<ev_cleanup> - even the best things end
		3410
		3411	Cleanup watchers are called just before the event loop is being destroyed
		3412	by a call to C<ev_loop_destroy>.
		3413
		3414	While there is no guarantee that the event loop gets destroyed, cleanup
		3415	watchers provide a convenient method to install cleanup hooks for your
		3416	program, worker threads and so on - you just to make sure to destroy the
		3417	loop when you want them to be invoked.
		3418
		3419	Cleanup watchers are invoked in the same way as any other watcher. Unlike
		3420	all other watchers, they do not keep a reference to the event loop (which
		3421	makes a lot of sense if you think about it). Like all other watchers, you
		3422	can call libev functions in the callback, except C<ev_cleanup_start>.
		3423
		3424	=head3 Watcher-Specific Functions and Data Members
		3425
		3426	=over 4
		3427
		3428	=item ev_cleanup_init (ev_cleanup *, callback)
		3429
		3430	Initialises and configures the cleanup watcher - it has no parameters of
		3431	any kind. There is a C<ev_cleanup_set> macro, but using it is utterly
		3432	pointless, I assure you.
		3433
		3434	=back
		3435
		3436	Example: Register an atexit handler to destroy the default loop, so any
		3437	cleanup functions are called.
		3438
		3439	static void
		3440	program_exits (void)
		3441	{
		3442	ev_loop_destroy (EV_DEFAULT_UC);
		3443	}
		3444
		3445	...
		3446	atexit (program_exits);
		3447
		3448
		3449	=head2 C<ev_async> - how to wake up an event loop
		3450
		3451	In general, you cannot use an C<ev_loop> from multiple threads or other
		3452	asynchronous sources such as signal handlers (as opposed to multiple event
		3453	loops - those are of course safe to use in different threads).
		3454
		3455	Sometimes, however, you need to wake up an event loop you do not control,
		3456	for example because it belongs to another thread. This is what C<ev_async>
		3457	watchers do: as long as the C<ev_async> watcher is active, you can signal
		3458	it by calling C<ev_async_send>, which is thread- and signal safe.
		3459
		3460	This functionality is very similar to C<ev_signal> watchers, as signals,
		3461	too, are asynchronous in nature, and signals, too, will be compressed
		3462	(i.e. the number of callback invocations may be less than the number of
		3463	C<ev_async_send> calls). In fact, you could use signal watchers as a kind
		3464	of "global async watchers" by using a watcher on an otherwise unused
		3465	signal, and C<ev_feed_signal> to signal this watcher from another thread,
		3466	even without knowing which loop owns the signal.
		3467
		3468	=head3 Queueing
		3469
		3470	C<ev_async> does not support queueing of data in any way. The reason
		3471	is that the author does not know of a simple (or any) algorithm for a
		3472	multiple-writer-single-reader queue that works in all cases and doesn't
		3473	need elaborate support such as pthreads or unportable memory access
		3474	semantics.
		3475
		3476	That means that if you want to queue data, you have to provide your own
		3477	queue. But at least I can tell you how to implement locking around your
		3478	queue:
		3479
		3480	=over 4
		3481
		3482	=item queueing from a signal handler context
		3483
		3484	To implement race-free queueing, you simply add to the queue in the signal
		3485	handler but you block the signal handler in the watcher callback. Here is
		3486	an example that does that for some fictitious SIGUSR1 handler:
		3487
		3488	static ev_async mysig;
		3489
		3490	static void
		3491	sigusr1_handler (void)
		3492	{
		3493	sometype data;
		3494
		3495	// no locking etc.
		3496	queue_put (data);
		3497	ev_async_send (EV_DEFAULT_ &mysig);
		3498	}
		3499
		3500	static void
		3501	mysig_cb (EV_P_ ev_async *w, int revents)
		3502	{
		3503	sometype data;
		3504	sigset_t block, prev;
		3505
		3506	sigemptyset (&block);
		3507	sigaddset (&block, SIGUSR1);
		3508	sigprocmask (SIG_BLOCK, &block, &prev);
		3509
		3510	while (queue_get (&data))
		3511	process (data);
		3512
		3513	if (sigismember (&prev, SIGUSR1)
		3514	sigprocmask (SIG_UNBLOCK, &block, 0);
		3515	}
		3516
		3517	(Note: pthreads in theory requires you to use C<pthread_setmask>
		3518	instead of C<sigprocmask> when you use threads, but libev doesn't do it
		3519	either...).
		3520
		3521	=item queueing from a thread context
		3522
		3523	The strategy for threads is different, as you cannot (easily) block
		3524	threads but you can easily preempt them, so to queue safely you need to
		3525	employ a traditional mutex lock, such as in this pthread example:
		3526
		3527	static ev_async mysig;
		3528	static pthread_mutex_t mymutex = PTHREAD_MUTEX_INITIALIZER;
		3529
		3530	static void
		3531	otherthread (void)
		3532	{
		3533	// only need to lock the actual queueing operation
		3534	pthread_mutex_lock (&mymutex);
		3535	queue_put (data);
		3536	pthread_mutex_unlock (&mymutex);
		3537
		3538	ev_async_send (EV_DEFAULT_ &mysig);
		3539	}
		3540
		3541	static void
		3542	mysig_cb (EV_P_ ev_async *w, int revents)
		3543	{
		3544	pthread_mutex_lock (&mymutex);
		3545
		3546	while (queue_get (&data))
		3547	process (data);
		3548
		3549	pthread_mutex_unlock (&mymutex);
		3550	}
		3551
		3552	=back
		3553
		3554
		3555	=head3 Watcher-Specific Functions and Data Members
		3556
		3557	=over 4
		3558
		3559	=item ev_async_init (ev_async *, callback)
		3560
		3561	Initialises and configures the async watcher - it has no parameters of any
		3562	kind. There is a C<ev_async_set> macro, but using it is utterly pointless,
		3563	trust me.
		3564
		3565	=item ev_async_send (loop, ev_async *)
		3566
		3567	Sends/signals/activates the given C<ev_async> watcher, that is, feeds
		3568	an C<EV_ASYNC> event on the watcher into the event loop, and instantly
		3569	returns.
		3570
		3571	Unlike C<ev_feed_event>, this call is safe to do from other threads,
		3572	signal or similar contexts (see the discussion of C<EV_ATOMIC_T> in the
		3573	embedding section below on what exactly this means).
		3574
		3575	Note that, as with other watchers in libev, multiple events might get
		3576	compressed into a single callback invocation (another way to look at
		3577	this is that C<ev_async> watchers are level-triggered: they are set on
		3578	C<ev_async_send>, reset when the event loop detects that).
		3579
		3580	This call incurs the overhead of at most one extra system call per event
		3581	loop iteration, if the event loop is blocked, and no syscall at all if
		3582	the event loop (or your program) is processing events. That means that
		3583	repeated calls are basically free (there is no need to avoid calls for
		3584	performance reasons) and that the overhead becomes smaller (typically
		3585	zero) under load.
		3586
		3587	=item bool = ev_async_pending (ev_async *)
		3588
		3589	Returns a non-zero value when C<ev_async_send> has been called on the
		3590	watcher but the event has not yet been processed (or even noted) by the
		3591	event loop.
		3592
		3593	C<ev_async_send> sets a flag in the watcher and wakes up the loop. When
		3594	the loop iterates next and checks for the watcher to have become active,
		3595	it will reset the flag again. C<ev_async_pending> can be used to very
		3596	quickly check whether invoking the loop might be a good idea.
		3597
		3598	Not that this does I<not> check whether the watcher itself is pending,
		3599	only whether it has been requested to make this watcher pending: there
		3600	is a time window between the event loop checking and resetting the async
		3601	notification, and the callback being invoked.
		3602
		3603	=back
		3604
		3605
740	=head1 OTHER FUNCTIONS	3606	=head1 OTHER FUNCTIONS
741		3607
742	There are some other functions of possible interest. Described. Here. Now.	3608	There are some other functions of possible interest. Described. Here. Now.
743		3609
744	=over 4	3610	=over 4
745		3611
746	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback)	3612	=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback, arg)
747		3613
748	This function combines a simple timer and an I/O watcher, calls your	3614	This function combines a simple timer and an I/O watcher, calls your
749	callback on whichever event happens first and automatically stop both	3615	callback on whichever event happens first and automatically stops both
750	watchers. This is useful if you want to wait for a single event on an fd	3616	watchers. This is useful if you want to wait for a single event on an fd
751	or timeout without having to allocate/configure/start/stop/free one or	3617	or timeout without having to allocate/configure/start/stop/free one or
752	more watchers yourself.	3618	more watchers yourself.
753		3619
754	If C<fd> is less than 0, then no I/O watcher will be started and events	3620	If C<fd> is less than 0, then no I/O watcher will be started and the
755	is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and	3621	C<events> argument is being ignored. Otherwise, an C<ev_io> watcher for
756	C<events> set will be craeted and started.	3622	the given C<fd> and C<events> set will be created and started.
757		3623
758	If C<timeout> is less than 0, then no timeout watcher will be	3624	If C<timeout> is less than 0, then no timeout watcher will be
759	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and	3625	started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and
760	repeat = 0) will be started. While C<0> is a valid timeout, it is of	3626	repeat = 0) will be started. C<0> is a valid timeout.
761	dubious value.
762		3627
763	The callback has the type C<void (*cb)(int revents, void *arg)> and gets	3628	The callback has the type C<void (*cb)(int revents, void *arg)> and is
764	passed an C<revents> set like normal event callbacks (a combination of	3629	passed an C<revents> set like normal event callbacks (a combination of
765	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg>	3630	C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMER>) and the C<arg>
766	value passed to C<ev_once>:	3631	value passed to C<ev_once>. Note that it is possible to receive I<both>
		3632	a timeout and an io event at the same time - you probably should give io
		3633	events precedence.
767		3634
		3635	Example: wait up to ten seconds for data to appear on STDIN_FILENO.
		3636
768	static void stdin_ready (int revents, void *arg)	3637	static void stdin_ready (int revents, void *arg)
		3638	{
		3639	if (revents & EV_READ)
		3640	/* stdin might have data for us, joy! */;
		3641	else if (revents & EV_TIMER)
		3642	/* doh, nothing entered */;
		3643	}
		3644
		3645	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);
		3646
		3647	=item ev_feed_fd_event (loop, int fd, int revents)
		3648
		3649	Feed an event on the given fd, as if a file descriptor backend detected
		3650	the given events.
		3651
		3652	=item ev_feed_signal_event (loop, int signum)
		3653
		3654	Feed an event as if the given signal occurred. See also C<ev_feed_signal>,
		3655	which is async-safe.
		3656
		3657	=back
		3658
		3659
		3660	=head1 COMMON OR USEFUL IDIOMS (OR BOTH)
		3661
		3662	This section explains some common idioms that are not immediately
		3663	obvious. Note that examples are sprinkled over the whole manual, and this
		3664	section only contains stuff that wouldn't fit anywhere else.
		3665
		3666	=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER
		3667
		3668	Each watcher has, by default, a C<void *data> member that you can read
		3669	or modify at any time: libev will completely ignore it. This can be used
		3670	to associate arbitrary data with your watcher. If you need more data and
		3671	don't want to allocate memory separately and store a pointer to it in that
		3672	data member, you can also "subclass" the watcher type and provide your own
		3673	data:
		3674
		3675	struct my_io
		3676	{
		3677	ev_io io;
		3678	int otherfd;
		3679	void *somedata;
		3680	struct whatever *mostinteresting;
		3681	};
		3682
		3683	...
		3684	struct my_io w;
		3685	ev_io_init (&w.io, my_cb, fd, EV_READ);
		3686
		3687	And since your callback will be called with a pointer to the watcher, you
		3688	can cast it back to your own type:
		3689
		3690	static void my_cb (struct ev_loop *loop, ev_io *w_, int revents)
		3691	{
		3692	struct my_io *w = (struct my_io *)w_;
		3693	...
		3694	}
		3695
		3696	More interesting and less C-conformant ways of casting your callback
		3697	function type instead have been omitted.
		3698
		3699	=head2 BUILDING YOUR OWN COMPOSITE WATCHERS
		3700
		3701	Another common scenario is to use some data structure with multiple
		3702	embedded watchers, in effect creating your own watcher that combines
		3703	multiple libev event sources into one "super-watcher":
		3704
		3705	struct my_biggy
		3706	{
		3707	int some_data;
		3708	ev_timer t1;
		3709	ev_timer t2;
		3710	}
		3711
		3712	In this case getting the pointer to C<my_biggy> is a bit more
		3713	complicated: Either you store the address of your C<my_biggy> struct in
		3714	the C<data> member of the watcher (for woozies or C++ coders), or you need
		3715	to use some pointer arithmetic using C<offsetof> inside your watchers (for
		3716	real programmers):
		3717
		3718	#include <stddef.h>
		3719
		3720	static void
		3721	t1_cb (EV_P_ ev_timer *w, int revents)
		3722	{
		3723	struct my_biggy big = (struct my_biggy *)
		3724	(((char *)w) - offsetof (struct my_biggy, t1));
		3725	}
		3726
		3727	static void
		3728	t2_cb (EV_P_ ev_timer *w, int revents)
		3729	{
		3730	struct my_biggy big = (struct my_biggy *)
		3731	(((char *)w) - offsetof (struct my_biggy, t2));
		3732	}
		3733
		3734	=head2 AVOIDING FINISHING BEFORE RETURNING
		3735
		3736	Often you have structures like this in event-based programs:
		3737
		3738	callback ()
769	{	3739	{
770	if (revents & EV_TIMEOUT)	3740	free (request);
771	/* doh, nothing entered */;
772	else if (revents & EV_READ)
773	/* stdin might have data for us, joy! */;
774	}	3741	}
775		3742
776	ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);	3743	request = start_new_request (..., callback);
777		3744
778	=item ev_feed_event (loop, watcher, int events)	3745	The intent is to start some "lengthy" operation. The C<request> could be
		3746	used to cancel the operation, or do other things with it.
779		3747
780	Feeds the given event set into the event loop, as if the specified event	3748	It's not uncommon to have code paths in C<start_new_request> that
781	had happened for the specified watcher (which must be a pointer to an	3749	immediately invoke the callback, for example, to report errors. Or you add
782	initialised but not necessarily started event watcher).	3750	some caching layer that finds that it can skip the lengthy aspects of the
		3751	operation and simply invoke the callback with the result.
783		3752
784	=item ev_feed_fd_event (loop, int fd, int revents)	3753	The problem here is that this will happen I<before> C<start_new_request>
		3754	has returned, so C<request> is not set.
785		3755
786	Feed an event on the given fd, as if a file descriptor backend detected	3756	Even if you pass the request by some safer means to the callback, you
787	the given events it.	3757	might want to do something to the request after starting it, such as
		3758	canceling it, which probably isn't working so well when the callback has
		3759	already been invoked.
788		3760
789	=item ev_feed_signal_event (loop, int signum)	3761	A common way around all these issues is to make sure that
		3762	C<start_new_request> I<always> returns before the callback is invoked. If
		3763	C<start_new_request> immediately knows the result, it can artificially
		3764	delay invoking the callback by using a C<prepare> or C<idle> watcher for
		3765	example, or more sneakily, by reusing an existing (stopped) watcher and
		3766	pushing it into the pending queue:
790		3767
791	Feed an event as if the given signal occured (loop must be the default loop!).	3768	ev_set_cb (watcher, callback);
		3769	ev_feed_event (EV_A_ watcher, 0);
792		3770
793	=back	3771	This way, C<start_new_request> can safely return before the callback is
		3772	invoked, while not delaying callback invocation too much.
		3773
		3774	=head2 MODEL/NESTED EVENT LOOP INVOCATIONS AND EXIT CONDITIONS
		3775
		3776	Often (especially in GUI toolkits) there are places where you have
		3777	I<modal> interaction, which is most easily implemented by recursively
		3778	invoking C<ev_run>.
		3779
		3780	This brings the problem of exiting - a callback might want to finish the
		3781	main C<ev_run> call, but not the nested one (e.g. user clicked "Quit", but
		3782	a modal "Are you sure?" dialog is still waiting), or just the nested one
		3783	and not the main one (e.g. user clocked "Ok" in a modal dialog), or some
		3784	other combination: In these cases, a simple C<ev_break> will not work.
		3785
		3786	The solution is to maintain "break this loop" variable for each C<ev_run>
		3787	invocation, and use a loop around C<ev_run> until the condition is
		3788	triggered, using C<EVRUN_ONCE>:
		3789
		3790	// main loop
		3791	int exit_main_loop = 0;
		3792
		3793	while (!exit_main_loop)
		3794	ev_run (EV_DEFAULT_ EVRUN_ONCE);
		3795
		3796	// in a modal watcher
		3797	int exit_nested_loop = 0;
		3798
		3799	while (!exit_nested_loop)
		3800	ev_run (EV_A_ EVRUN_ONCE);
		3801
		3802	To exit from any of these loops, just set the corresponding exit variable:
		3803
		3804	// exit modal loop
		3805	exit_nested_loop = 1;
		3806
		3807	// exit main program, after modal loop is finished
		3808	exit_main_loop = 1;
		3809
		3810	// exit both
		3811	exit_main_loop = exit_nested_loop = 1;
		3812
		3813	=head2 THREAD LOCKING EXAMPLE
		3814
		3815	Here is a fictitious example of how to run an event loop in a different
		3816	thread from where callbacks are being invoked and watchers are
		3817	created/added/removed.
		3818
		3819	For a real-world example, see the C<EV::Loop::Async> perl module,
		3820	which uses exactly this technique (which is suited for many high-level
		3821	languages).
		3822
		3823	The example uses a pthread mutex to protect the loop data, a condition
		3824	variable to wait for callback invocations, an async watcher to notify the
		3825	event loop thread and an unspecified mechanism to wake up the main thread.
		3826
		3827	First, you need to associate some data with the event loop:
		3828
		3829	typedef struct {
		3830	mutex_t lock; /* global loop lock */
		3831	ev_async async_w;
		3832	thread_t tid;
		3833	cond_t invoke_cv;
		3834	} userdata;
		3835
		3836	void prepare_loop (EV_P)
		3837	{
		3838	// for simplicity, we use a static userdata struct.
		3839	static userdata u;
		3840
		3841	ev_async_init (&u->async_w, async_cb);
		3842	ev_async_start (EV_A_ &u->async_w);
		3843
		3844	pthread_mutex_init (&u->lock, 0);
		3845	pthread_cond_init (&u->invoke_cv, 0);
		3846
		3847	// now associate this with the loop
		3848	ev_set_userdata (EV_A_ u);
		3849	ev_set_invoke_pending_cb (EV_A_ l_invoke);
		3850	ev_set_loop_release_cb (EV_A_ l_release, l_acquire);
		3851
		3852	// then create the thread running ev_run
		3853	pthread_create (&u->tid, 0, l_run, EV_A);
		3854	}
		3855
		3856	The callback for the C<ev_async> watcher does nothing: the watcher is used
		3857	solely to wake up the event loop so it takes notice of any new watchers
		3858	that might have been added:
		3859
		3860	static void
		3861	async_cb (EV_P_ ev_async *w, int revents)
		3862	{
		3863	// just used for the side effects
		3864	}
		3865
		3866	The C<l_release> and C<l_acquire> callbacks simply unlock/lock the mutex
		3867	protecting the loop data, respectively.
		3868
		3869	static void
		3870	l_release (EV_P)
		3871	{
		3872	userdata *u = ev_userdata (EV_A);
		3873	pthread_mutex_unlock (&u->lock);
		3874	}
		3875
		3876	static void
		3877	l_acquire (EV_P)
		3878	{
		3879	userdata *u = ev_userdata (EV_A);
		3880	pthread_mutex_lock (&u->lock);
		3881	}
		3882
		3883	The event loop thread first acquires the mutex, and then jumps straight
		3884	into C<ev_run>:
		3885
		3886	void *
		3887	l_run (void *thr_arg)
		3888	{
		3889	struct ev_loop *loop = (struct ev_loop *)thr_arg;
		3890
		3891	l_acquire (EV_A);
		3892	pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
		3893	ev_run (EV_A_ 0);
		3894	l_release (EV_A);
		3895
		3896	return 0;
		3897	}
		3898
		3899	Instead of invoking all pending watchers, the C<l_invoke> callback will
		3900	signal the main thread via some unspecified mechanism (signals? pipe
		3901	writes? C<Async::Interrupt>?) and then waits until all pending watchers
		3902	have been called (in a while loop because a) spurious wakeups are possible
		3903	and b) skipping inter-thread-communication when there are no pending
		3904	watchers is very beneficial):
		3905
		3906	static void
		3907	l_invoke (EV_P)
		3908	{
		3909	userdata *u = ev_userdata (EV_A);
		3910
		3911	while (ev_pending_count (EV_A))
		3912	{
		3913	wake_up_other_thread_in_some_magic_or_not_so_magic_way ();
		3914	pthread_cond_wait (&u->invoke_cv, &u->lock);
		3915	}
		3916	}
		3917
		3918	Now, whenever the main thread gets told to invoke pending watchers, it
		3919	will grab the lock, call C<ev_invoke_pending> and then signal the loop
		3920	thread to continue:
		3921
		3922	static void
		3923	real_invoke_pending (EV_P)
		3924	{
		3925	userdata *u = ev_userdata (EV_A);
		3926
		3927	pthread_mutex_lock (&u->lock);
		3928	ev_invoke_pending (EV_A);
		3929	pthread_cond_signal (&u->invoke_cv);
		3930	pthread_mutex_unlock (&u->lock);
		3931	}
		3932
		3933	Whenever you want to start/stop a watcher or do other modifications to an
		3934	event loop, you will now have to lock:
		3935
		3936	ev_timer timeout_watcher;
		3937	userdata *u = ev_userdata (EV_A);
		3938
		3939	ev_timer_init (&timeout_watcher, timeout_cb, 5.5, 0.);
		3940
		3941	pthread_mutex_lock (&u->lock);
		3942	ev_timer_start (EV_A_ &timeout_watcher);
		3943	ev_async_send (EV_A_ &u->async_w);
		3944	pthread_mutex_unlock (&u->lock);
		3945
		3946	Note that sending the C<ev_async> watcher is required because otherwise
		3947	an event loop currently blocking in the kernel will have no knowledge
		3948	about the newly added timer. By waking up the loop it will pick up any new
		3949	watchers in the next event loop iteration.
		3950
		3951	=head2 THREADS, COROUTINES, CONTINUATIONS, QUEUES... INSTEAD OF CALLBACKS
		3952
		3953	While the overhead of a callback that e.g. schedules a thread is small, it
		3954	is still an overhead. If you embed libev, and your main usage is with some
		3955	kind of threads or coroutines, you might want to customise libev so that
		3956	doesn't need callbacks anymore.
		3957
		3958	Imagine you have coroutines that you can switch to using a function
		3959	C<switch_to (coro)>, that libev runs in a coroutine called C<libev_coro>
		3960	and that due to some magic, the currently active coroutine is stored in a
		3961	global called C<current_coro>. Then you can build your own "wait for libev
		3962	event" primitive by changing C<EV_CB_DECLARE> and C<EV_CB_INVOKE> (note
		3963	the differing C<;> conventions):
		3964
		3965	#define EV_CB_DECLARE(type) struct my_coro *cb;
		3966	#define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb)
		3967
		3968	That means instead of having a C callback function, you store the
		3969	coroutine to switch to in each watcher, and instead of having libev call
		3970	your callback, you instead have it switch to that coroutine.
		3971
		3972	A coroutine might now wait for an event with a function called
		3973	C<wait_for_event>. (the watcher needs to be started, as always, but it doesn't
		3974	matter when, or whether the watcher is active or not when this function is
		3975	called):
		3976
		3977	void
		3978	wait_for_event (ev_watcher *w)
		3979	{
		3980	ev_set_cb (w, current_coro);
		3981	switch_to (libev_coro);
		3982	}
		3983
		3984	That basically suspends the coroutine inside C<wait_for_event> and
		3985	continues the libev coroutine, which, when appropriate, switches back to
		3986	this or any other coroutine.
		3987
		3988	You can do similar tricks if you have, say, threads with an event queue -
		3989	instead of storing a coroutine, you store the queue object and instead of
		3990	switching to a coroutine, you push the watcher onto the queue and notify
		3991	any waiters.
		3992
		3993	To embed libev, see L</EMBEDDING>, but in short, it's easiest to create two
		3994	files, F<my_ev.h> and F<my_ev.c> that include the respective libev files:
		3995
		3996	// my_ev.h
		3997	#define EV_CB_DECLARE(type) struct my_coro *cb;
		3998	#define EV_CB_INVOKE(watcher) switch_to ((watcher)->cb)
		3999	#include "../libev/ev.h"
		4000
		4001	// my_ev.c
		4002	#define EV_H "my_ev.h"
		4003	#include "../libev/ev.c"
		4004
		4005	And then use F<my_ev.h> when you would normally use F<ev.h>, and compile
		4006	F<my_ev.c> into your project. When properly specifying include paths, you
		4007	can even use F<ev.h> as header file name directly.
		4008
794		4009
795	=head1 LIBEVENT EMULATION	4010	=head1 LIBEVENT EMULATION
796		4011
797	Libev offers a compatibility emulation layer for libevent. It cannot	4012	Libev offers a compatibility emulation layer for libevent. It cannot
798	emulate the internals of libevent, so here are some usage hints:	4013	emulate the internals of libevent, so here are some usage hints:
799		4014
800	=over 4	4015	=over 4
		4016
		4017	=item * Only the libevent-1.4.1-beta API is being emulated.
		4018
		4019	This was the newest libevent version available when libev was implemented,
		4020	and is still mostly unchanged in 2010.
801		4021
802	=item * Use it by including <event.h>, as usual.	4022	=item * Use it by including <event.h>, as usual.
803		4023
804	=item * The following members are fully supported: ev_base, ev_callback,	4024	=item * The following members are fully supported: ev_base, ev_callback,
805	ev_arg, ev_fd, ev_res, ev_events.	4025	ev_arg, ev_fd, ev_res, ev_events.
…		…
810		4030
811	=item * Priorities are not currently supported. Initialising priorities	4031	=item * Priorities are not currently supported. Initialising priorities
812	will fail and all watchers will have the same priority, even though there	4032	will fail and all watchers will have the same priority, even though there
813	is an ev_pri field.	4033	is an ev_pri field.
814		4034
		4035	=item * In libevent, the last base created gets the signals, in libev, the
		4036	base that registered the signal gets the signals.
		4037
815	=item * Other members are not supported.	4038	=item * Other members are not supported.
816		4039
817	=item * The libev emulation is I<not> ABI compatible to libevent, you need	4040	=item * The libev emulation is I<not> ABI compatible to libevent, you need
818	to use the libev header file and library.	4041	to use the libev header file and library.
819		4042
820	=back	4043	=back
821		4044
822	=head1 C++ SUPPORT	4045	=head1 C++ SUPPORT
823		4046
824	TBD.	4047	=head2 C API
		4048
		4049	The normal C API should work fine when used from C++: both ev.h and the
		4050	libev sources can be compiled as C++. Therefore, code that uses the C API
		4051	will work fine.
		4052
		4053	Proper exception specifications might have to be added to callbacks passed
		4054	to libev: exceptions may be thrown only from watcher callbacks, all other
		4055	callbacks (allocator, syserr, loop acquire/release and periodic reschedule
		4056	callbacks) must not throw exceptions, and might need a C<noexcept>
		4057	specification. If you have code that needs to be compiled as both C and
		4058	C++ you can use the C<EV_NOEXCEPT> macro for this:
		4059
		4060	static void
		4061	fatal_error (const char *msg) EV_NOEXCEPT
		4062	{
		4063	perror (msg);
		4064	abort ();
		4065	}
		4066
		4067	...
		4068	ev_set_syserr_cb (fatal_error);
		4069
		4070	The only API functions that can currently throw exceptions are C<ev_run>,
		4071	C<ev_invoke>, C<ev_invoke_pending> and C<ev_loop_destroy> (the latter
		4072	because it runs cleanup watchers).
		4073
		4074	Throwing exceptions in watcher callbacks is only supported if libev itself
		4075	is compiled with a C++ compiler or your C and C++ environments allow
		4076	throwing exceptions through C libraries (most do).
		4077
		4078	=head2 C++ API
		4079
		4080	Libev comes with some simplistic wrapper classes for C++ that mainly allow
		4081	you to use some convenience methods to start/stop watchers and also change
		4082	the callback model to a model using method callbacks on objects.
		4083
		4084	To use it,
		4085
		4086	#include <ev++.h>
		4087
		4088	This automatically includes F<ev.h> and puts all of its definitions (many
		4089	of them macros) into the global namespace. All C++ specific things are
		4090	put into the C<ev> namespace. It should support all the same embedding
		4091	options as F<ev.h>, most notably C<EV_MULTIPLICITY>.
		4092
		4093	Care has been taken to keep the overhead low. The only data member the C++
		4094	classes add (compared to plain C-style watchers) is the event loop pointer
		4095	that the watcher is associated with (or no additional members at all if
		4096	you disable C<EV_MULTIPLICITY> when embedding libev).
		4097
		4098	Currently, functions, static and non-static member functions and classes
		4099	with C<operator ()> can be used as callbacks. Other types should be easy
		4100	to add as long as they only need one additional pointer for context. If
		4101	you need support for other types of functors please contact the author
		4102	(preferably after implementing it).
		4103
		4104	For all this to work, your C++ compiler either has to use the same calling
		4105	conventions as your C compiler (for static member functions), or you have
		4106	to embed libev and compile libev itself as C++.
		4107
		4108	Here is a list of things available in the C<ev> namespace:
		4109
		4110	=over 4
		4111
		4112	=item C<ev::READ>, C<ev::WRITE> etc.
		4113
		4114	These are just enum values with the same values as the C<EV_READ> etc.
		4115	macros from F<ev.h>.
		4116
		4117	=item C<ev::tstamp>, C<ev::now>
		4118
		4119	Aliases to the same types/functions as with the C<ev_> prefix.
		4120
		4121	=item C<ev::io>, C<ev::timer>, C<ev::periodic>, C<ev::idle>, C<ev::sig> etc.
		4122
		4123	For each C<ev_TYPE> watcher in F<ev.h> there is a corresponding class of
		4124	the same name in the C<ev> namespace, with the exception of C<ev_signal>
		4125	which is called C<ev::sig> to avoid clashes with the C<signal> macro
		4126	defined by many implementations.
		4127
		4128	All of those classes have these methods:
		4129
		4130	=over 4
		4131
		4132	=item ev::TYPE::TYPE ()
		4133
		4134	=item ev::TYPE::TYPE (loop)
		4135
		4136	=item ev::TYPE::~TYPE
		4137
		4138	The constructor (optionally) takes an event loop to associate the watcher
		4139	with. If it is omitted, it will use C<EV_DEFAULT>.
		4140
		4141	The constructor calls C<ev_init> for you, which means you have to call the
		4142	C<set> method before starting it.
		4143
		4144	It will not set a callback, however: You have to call the templated C<set>
		4145	method to set a callback before you can start the watcher.
		4146
		4147	(The reason why you have to use a method is a limitation in C++ which does
		4148	not allow explicit template arguments for constructors).
		4149
		4150	The destructor automatically stops the watcher if it is active.
		4151
		4152	=item w->set<class, &class::method> (object *)
		4153
		4154	This method sets the callback method to call. The method has to have a
		4155	signature of C<void (*)(ev_TYPE &, int)>, it receives the watcher as
		4156	first argument and the C<revents> as second. The object must be given as
		4157	parameter and is stored in the C<data> member of the watcher.
		4158
		4159	This method synthesizes efficient thunking code to call your method from
		4160	the C callback that libev requires. If your compiler can inline your
		4161	callback (i.e. it is visible to it at the place of the C<set> call and
		4162	your compiler is good :), then the method will be fully inlined into the
		4163	thunking function, making it as fast as a direct C callback.
		4164
		4165	Example: simple class declaration and watcher initialisation
		4166
		4167	struct myclass
		4168	{
		4169	void io_cb (ev::io &w, int revents) { }
		4170	}
		4171
		4172	myclass obj;
		4173	ev::io iow;
		4174	iow.set <myclass, &myclass::io_cb> (&obj);
		4175
		4176	=item w->set (object *)
		4177
		4178	This is a variation of a method callback - leaving out the method to call
		4179	will default the method to C<operator ()>, which makes it possible to use
		4180	functor objects without having to manually specify the C<operator ()> all
		4181	the time. Incidentally, you can then also leave out the template argument
		4182	list.
		4183
		4184	The C<operator ()> method prototype must be C<void operator ()(watcher &w,
		4185	int revents)>.
		4186
		4187	See the method-C<set> above for more details.
		4188
		4189	Example: use a functor object as callback.
		4190
		4191	struct myfunctor
		4192	{
		4193	void operator() (ev::io &w, int revents)
		4194	{
		4195	...
		4196	}
		4197	}
		4198
		4199	myfunctor f;
		4200
		4201	ev::io w;
		4202	w.set (&f);
		4203
		4204	=item w->set<function> (void *data = 0)
		4205
		4206	Also sets a callback, but uses a static method or plain function as
		4207	callback. The optional C<data> argument will be stored in the watcher's
		4208	C<data> member and is free for you to use.
		4209
		4210	The prototype of the C<function> must be C<void (*)(ev::TYPE &w, int)>.
		4211
		4212	See the method-C<set> above for more details.
		4213
		4214	Example: Use a plain function as callback.
		4215
		4216	static void io_cb (ev::io &w, int revents) { }
		4217	iow.set <io_cb> ();
		4218
		4219	=item w->set (loop)
		4220
		4221	Associates a different C<struct ev_loop> with this watcher. You can only
		4222	do this when the watcher is inactive (and not pending either).
		4223
		4224	=item w->set ([arguments])
		4225
		4226	Basically the same as C<ev_TYPE_set> (except for C<ev::embed> watchers>),
		4227	with the same arguments. Either this method or a suitable start method
		4228	must be called at least once. Unlike the C counterpart, an active watcher
		4229	gets automatically stopped and restarted when reconfiguring it with this
		4230	method.
		4231
		4232	For C<ev::embed> watchers this method is called C<set_embed>, to avoid
		4233	clashing with the C<set (loop)> method.
		4234
		4235	=item w->start ()
		4236
		4237	Starts the watcher. Note that there is no C<loop> argument, as the
		4238	constructor already stores the event loop.
		4239
		4240	=item w->start ([arguments])
		4241
		4242	Instead of calling C<set> and C<start> methods separately, it is often
		4243	convenient to wrap them in one call. Uses the same type of arguments as
		4244	the configure C<set> method of the watcher.
		4245
		4246	=item w->stop ()
		4247
		4248	Stops the watcher if it is active. Again, no C<loop> argument.
		4249
		4250	=item w->again () (C<ev::timer>, C<ev::periodic> only)
		4251
		4252	For C<ev::timer> and C<ev::periodic>, this invokes the corresponding
		4253	C<ev_TYPE_again> function.
		4254
		4255	=item w->sweep () (C<ev::embed> only)
		4256
		4257	Invokes C<ev_embed_sweep>.
		4258
		4259	=item w->update () (C<ev::stat> only)
		4260
		4261	Invokes C<ev_stat_stat>.
		4262
		4263	=back
		4264
		4265	=back
		4266
		4267	Example: Define a class with two I/O and idle watchers, start the I/O
		4268	watchers in the constructor.
		4269
		4270	class myclass
		4271	{
		4272	ev::io io ; void io_cb (ev::io &w, int revents);
		4273	ev::io io2 ; void io2_cb (ev::io &w, int revents);
		4274	ev::idle idle; void idle_cb (ev::idle &w, int revents);
		4275
		4276	myclass (int fd)
		4277	{
		4278	io .set <myclass, &myclass::io_cb > (this);
		4279	io2 .set <myclass, &myclass::io2_cb > (this);
		4280	idle.set <myclass, &myclass::idle_cb> (this);
		4281
		4282	io.set (fd, ev::WRITE); // configure the watcher
		4283	io.start (); // start it whenever convenient
		4284
		4285	io2.start (fd, ev::READ); // set + start in one call
		4286	}
		4287	};
		4288
		4289
		4290	=head1 OTHER LANGUAGE BINDINGS
		4291
		4292	Libev does not offer other language bindings itself, but bindings for a
		4293	number of languages exist in the form of third-party packages. If you know
		4294	any interesting language binding in addition to the ones listed here, drop
		4295	me a note.
		4296
		4297	=over 4
		4298
		4299	=item Perl
		4300
		4301	The EV module implements the full libev API and is actually used to test
		4302	libev. EV is developed together with libev. Apart from the EV core module,
		4303	there are additional modules that implement libev-compatible interfaces
		4304	to C<libadns> (C<EV::ADNS>, but C<AnyEvent::DNS> is preferred nowadays),
		4305	C<Net::SNMP> (C<Net::SNMP::EV>) and the C<libglib> event core (C<Glib::EV>
		4306	and C<EV::Glib>).
		4307
		4308	It can be found and installed via CPAN, its homepage is at
		4309	L<http://software.schmorp.de/pkg/EV>.
		4310
		4311	=item Python
		4312
		4313	Python bindings can be found at L<http://code.google.com/p/pyev/>. It
		4314	seems to be quite complete and well-documented.
		4315
		4316	=item Ruby
		4317
		4318	Tony Arcieri has written a ruby extension that offers access to a subset
		4319	of the libev API and adds file handle abstractions, asynchronous DNS and
		4320	more on top of it. It can be found via gem servers. Its homepage is at
		4321	L<http://rev.rubyforge.org/>.
		4322
		4323	Roger Pack reports that using the link order C<-lws2_32 -lmsvcrt-ruby-190>
		4324	makes rev work even on mingw.
		4325
		4326	=item Haskell
		4327
		4328	A haskell binding to libev is available at
		4329	L<http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hlibev>.
		4330
		4331	=item D
		4332
		4333	Leandro Lucarella has written a D language binding (F<ev.d>) for libev, to
		4334	be found at L<http://www.llucax.com.ar/proj/ev.d/index.html>.
		4335
		4336	=item Ocaml
		4337
		4338	Erkki Seppala has written Ocaml bindings for libev, to be found at
		4339	L<http://modeemi.cs.tut.fi/~flux/software/ocaml-ev/>.
		4340
		4341	=item Lua
		4342
		4343	Brian Maher has written a partial interface to libev for lua (at the
		4344	time of this writing, only C<ev_io> and C<ev_timer>), to be found at
		4345	L<http://github.com/brimworks/lua-ev>.
		4346
		4347	=item Javascript
		4348
		4349	Node.js (L<http://nodejs.org>) uses libev as the underlying event library.
		4350
		4351	=item Others
		4352
		4353	There are others, and I stopped counting.
		4354
		4355	=back
		4356
		4357
		4358	=head1 MACRO MAGIC
		4359
		4360	Libev can be compiled with a variety of options, the most fundamental
		4361	of which is C<EV_MULTIPLICITY>. This option determines whether (most)
		4362	functions and callbacks have an initial C<struct ev_loop *> argument.
		4363
		4364	To make it easier to write programs that cope with either variant, the
		4365	following macros are defined:
		4366
		4367	=over 4
		4368
		4369	=item C<EV_A>, C<EV_A_>
		4370
		4371	This provides the loop I<argument> for functions, if one is required ("ev
		4372	loop argument"). The C<EV_A> form is used when this is the sole argument,
		4373	C<EV_A_> is used when other arguments are following. Example:
		4374
		4375	ev_unref (EV_A);
		4376	ev_timer_add (EV_A_ watcher);
		4377	ev_run (EV_A_ 0);
		4378
		4379	It assumes the variable C<loop> of type C<struct ev_loop *> is in scope,
		4380	which is often provided by the following macro.
		4381
		4382	=item C<EV_P>, C<EV_P_>
		4383
		4384	This provides the loop I<parameter> for functions, if one is required ("ev
		4385	loop parameter"). The C<EV_P> form is used when this is the sole parameter,
		4386	C<EV_P_> is used when other parameters are following. Example:
		4387
		4388	// this is how ev_unref is being declared
		4389	static void ev_unref (EV_P);
		4390
		4391	// this is how you can declare your typical callback
		4392	static void cb (EV_P_ ev_timer *w, int revents)
		4393
		4394	It declares a parameter C<loop> of type C<struct ev_loop *>, quite
		4395	suitable for use with C<EV_A>.
		4396
		4397	=item C<EV_DEFAULT>, C<EV_DEFAULT_>
		4398
		4399	Similar to the other two macros, this gives you the value of the default
		4400	loop, if multiple loops are supported ("ev loop default"). The default loop
		4401	will be initialised if it isn't already initialised.
		4402
		4403	For non-multiplicity builds, these macros do nothing, so you always have
		4404	to initialise the loop somewhere.
		4405
		4406	=item C<EV_DEFAULT_UC>, C<EV_DEFAULT_UC_>
		4407
		4408	Usage identical to C<EV_DEFAULT> and C<EV_DEFAULT_>, but requires that the
		4409	default loop has been initialised (C<UC> == unchecked). Their behaviour
		4410	is undefined when the default loop has not been initialised by a previous
		4411	execution of C<EV_DEFAULT>, C<EV_DEFAULT_> or C<ev_default_init (...)>.
		4412
		4413	It is often prudent to use C<EV_DEFAULT> when initialising the first
		4414	watcher in a function but use C<EV_DEFAULT_UC> afterwards.
		4415
		4416	=back
		4417
		4418	Example: Declare and initialise a check watcher, utilising the above
		4419	macros so it will work regardless of whether multiple loops are supported
		4420	or not.
		4421
		4422	static void
		4423	check_cb (EV_P_ ev_timer *w, int revents)
		4424	{
		4425	ev_check_stop (EV_A_ w);
		4426	}
		4427
		4428	ev_check check;
		4429	ev_check_init (&check, check_cb);
		4430	ev_check_start (EV_DEFAULT_ &check);
		4431	ev_run (EV_DEFAULT_ 0);
		4432
		4433	=head1 EMBEDDING
		4434
		4435	Libev can (and often is) directly embedded into host
		4436	applications. Examples of applications that embed it include the Deliantra
		4437	Game Server, the EV perl module, the GNU Virtual Private Ethernet (gvpe)
		4438	and rxvt-unicode.
		4439
		4440	The goal is to enable you to just copy the necessary files into your
		4441	source directory without having to change even a single line in them, so
		4442	you can easily upgrade by simply copying (or having a checked-out copy of
		4443	libev somewhere in your source tree).
		4444
		4445	=head2 FILESETS
		4446
		4447	Depending on what features you need you need to include one or more sets of files
		4448	in your application.
		4449
		4450	=head3 CORE EVENT LOOP
		4451
		4452	To include only the libev core (all the C<ev_*> functions), with manual
		4453	configuration (no autoconf):
		4454
		4455	#define EV_STANDALONE 1
		4456	#include "ev.c"
		4457
		4458	This will automatically include F<ev.h>, too, and should be done in a
		4459	single C source file only to provide the function implementations. To use
		4460	it, do the same for F<ev.h> in all files wishing to use this API (best
		4461	done by writing a wrapper around F<ev.h> that you can include instead and
		4462	where you can put other configuration options):
		4463
		4464	#define EV_STANDALONE 1
		4465	#include "ev.h"
		4466
		4467	Both header files and implementation files can be compiled with a C++
		4468	compiler (at least, that's a stated goal, and breakage will be treated
		4469	as a bug).
		4470
		4471	You need the following files in your source tree, or in a directory
		4472	in your include path (e.g. in libev/ when using -Ilibev):
		4473
		4474	ev.h
		4475	ev.c
		4476	ev_vars.h
		4477	ev_wrap.h
		4478
		4479	ev_win32.c required on win32 platforms only
		4480
		4481	ev_select.c only when select backend is enabled
		4482	ev_poll.c only when poll backend is enabled
		4483	ev_epoll.c only when the epoll backend is enabled
		4484	ev_linuxaio.c only when the linux aio backend is enabled
		4485	ev_kqueue.c only when the kqueue backend is enabled
		4486	ev_port.c only when the solaris port backend is enabled
		4487
		4488	F<ev.c> includes the backend files directly when enabled, so you only need
		4489	to compile this single file.
		4490
		4491	=head3 LIBEVENT COMPATIBILITY API
		4492
		4493	To include the libevent compatibility API, also include:
		4494
		4495	#include "event.c"
		4496
		4497	in the file including F<ev.c>, and:
		4498
		4499	#include "event.h"
		4500
		4501	in the files that want to use the libevent API. This also includes F<ev.h>.
		4502
		4503	You need the following additional files for this:
		4504
		4505	event.h
		4506	event.c
		4507
		4508	=head3 AUTOCONF SUPPORT
		4509
		4510	Instead of using C<EV_STANDALONE=1> and providing your configuration in
		4511	whatever way you want, you can also C<m4_include([libev.m4])> in your
		4512	F<configure.ac> and leave C<EV_STANDALONE> undefined. F<ev.c> will then
		4513	include F<config.h> and configure itself accordingly.
		4514
		4515	For this of course you need the m4 file:
		4516
		4517	libev.m4
		4518
		4519	=head2 PREPROCESSOR SYMBOLS/MACROS
		4520
		4521	Libev can be configured via a variety of preprocessor symbols you have to
		4522	define before including (or compiling) any of its files. The default in
		4523	the absence of autoconf is documented for every option.
		4524
		4525	Symbols marked with "(h)" do not change the ABI, and can have different
		4526	values when compiling libev vs. including F<ev.h>, so it is permissible
		4527	to redefine them before including F<ev.h> without breaking compatibility
		4528	to a compiled library. All other symbols change the ABI, which means all
		4529	users of libev and the libev code itself must be compiled with compatible
		4530	settings.
		4531
		4532	=over 4
		4533
		4534	=item EV_COMPAT3 (h)
		4535
		4536	Backwards compatibility is a major concern for libev. This is why this
		4537	release of libev comes with wrappers for the functions and symbols that
		4538	have been renamed between libev version 3 and 4.
		4539
		4540	You can disable these wrappers (to test compatibility with future
		4541	versions) by defining C<EV_COMPAT3> to C<0> when compiling your
		4542	sources. This has the additional advantage that you can drop the C<struct>
		4543	from C<struct ev_loop> declarations, as libev will provide an C<ev_loop>
		4544	typedef in that case.
		4545
		4546	In some future version, the default for C<EV_COMPAT3> will become C<0>,
		4547	and in some even more future version the compatibility code will be
		4548	removed completely.
		4549
		4550	=item EV_STANDALONE (h)
		4551
		4552	Must always be C<1> if you do not use autoconf configuration, which
		4553	keeps libev from including F<config.h>, and it also defines dummy
		4554	implementations for some libevent functions (such as logging, which is not
		4555	supported). It will also not define any of the structs usually found in
		4556	F<event.h> that are not directly supported by the libev core alone.
		4557
		4558	In standalone mode, libev will still try to automatically deduce the
		4559	configuration, but has to be more conservative.
		4560
		4561	=item EV_USE_FLOOR
		4562
		4563	If defined to be C<1>, libev will use the C<floor ()> function for its
		4564	periodic reschedule calculations, otherwise libev will fall back on a
		4565	portable (slower) implementation. If you enable this, you usually have to
		4566	link against libm or something equivalent. Enabling this when the C<floor>
		4567	function is not available will fail, so the safe default is to not enable
		4568	this.
		4569
		4570	=item EV_USE_MONOTONIC
		4571
		4572	If defined to be C<1>, libev will try to detect the availability of the
		4573	monotonic clock option at both compile time and runtime. Otherwise no
		4574	use of the monotonic clock option will be attempted. If you enable this,
		4575	you usually have to link against librt or something similar. Enabling it
		4576	when the functionality isn't available is safe, though, although you have
		4577	to make sure you link against any libraries where the C<clock_gettime>
		4578	function is hiding in (often F<-lrt>). See also C<EV_USE_CLOCK_SYSCALL>.
		4579
		4580	=item EV_USE_REALTIME
		4581
		4582	If defined to be C<1>, libev will try to detect the availability of the
		4583	real-time clock option at compile time (and assume its availability
		4584	at runtime if successful). Otherwise no use of the real-time clock
		4585	option will be attempted. This effectively replaces C<gettimeofday>
		4586	by C<clock_get (CLOCK_REALTIME, ...)> and will not normally affect
		4587	correctness. See the note about libraries in the description of
		4588	C<EV_USE_MONOTONIC>, though. Defaults to the opposite value of
		4589	C<EV_USE_CLOCK_SYSCALL>.
		4590
		4591	=item EV_USE_CLOCK_SYSCALL
		4592
		4593	If defined to be C<1>, libev will try to use a direct syscall instead
		4594	of calling the system-provided C<clock_gettime> function. This option
		4595	exists because on GNU/Linux, C<clock_gettime> is in C<librt>, but C<librt>
		4596	unconditionally pulls in C<libpthread>, slowing down single-threaded
		4597	programs needlessly. Using a direct syscall is slightly slower (in
		4598	theory), because no optimised vdso implementation can be used, but avoids
		4599	the pthread dependency. Defaults to C<1> on GNU/Linux with glibc 2.x or
		4600	higher, as it simplifies linking (no need for C<-lrt>).
		4601
		4602	=item EV_USE_NANOSLEEP
		4603
		4604	If defined to be C<1>, libev will assume that C<nanosleep ()> is available
		4605	and will use it for delays. Otherwise it will use C<select ()>.
		4606
		4607	=item EV_USE_EVENTFD
		4608
		4609	If defined to be C<1>, then libev will assume that C<eventfd ()> is
		4610	available and will probe for kernel support at runtime. This will improve
		4611	C<ev_signal> and C<ev_async> performance and reduce resource consumption.
		4612	If undefined, it will be enabled if the headers indicate GNU/Linux + Glibc
		4613	2.7 or newer, otherwise disabled.
		4614
		4615	=item EV_USE_SELECT
		4616
		4617	If undefined or defined to be C<1>, libev will compile in support for the
		4618	C<select>(2) backend. No attempt at auto-detection will be done: if no
		4619	other method takes over, select will be it. Otherwise the select backend
		4620	will not be compiled in.
		4621
		4622	=item EV_SELECT_USE_FD_SET
		4623
		4624	If defined to C<1>, then the select backend will use the system C<fd_set>
		4625	structure. This is useful if libev doesn't compile due to a missing
		4626	C<NFDBITS> or C<fd_mask> definition or it mis-guesses the bitset layout
		4627	on exotic systems. This usually limits the range of file descriptors to
		4628	some low limit such as 1024 or might have other limitations (winsocket
		4629	only allows 64 sockets). The C<FD_SETSIZE> macro, set before compilation,
		4630	configures the maximum size of the C<fd_set>.
		4631
		4632	=item EV_SELECT_IS_WINSOCKET
		4633
		4634	When defined to C<1>, the select backend will assume that
		4635	select/socket/connect etc. don't understand file descriptors but
		4636	wants osf handles on win32 (this is the case when the select to
		4637	be used is the winsock select). This means that it will call
		4638	C<_get_osfhandle> on the fd to convert it to an OS handle. Otherwise,
		4639	it is assumed that all these functions actually work on fds, even
		4640	on win32. Should not be defined on non-win32 platforms.
		4641
		4642	=item EV_FD_TO_WIN32_HANDLE(fd)
		4643
		4644	If C<EV_SELECT_IS_WINSOCKET> is enabled, then libev needs a way to map
		4645	file descriptors to socket handles. When not defining this symbol (the
		4646	default), then libev will call C<_get_osfhandle>, which is usually
		4647	correct. In some cases, programs use their own file descriptor management,
		4648	in which case they can provide this function to map fds to socket handles.
		4649
		4650	=item EV_WIN32_HANDLE_TO_FD(handle)
		4651
		4652	If C<EV_SELECT_IS_WINSOCKET> then libev maps handles to file descriptors
		4653	using the standard C<_open_osfhandle> function. For programs implementing
		4654	their own fd to handle mapping, overwriting this function makes it easier
		4655	to do so. This can be done by defining this macro to an appropriate value.
		4656
		4657	=item EV_WIN32_CLOSE_FD(fd)
		4658
		4659	If programs implement their own fd to handle mapping on win32, then this
		4660	macro can be used to override the C<close> function, useful to unregister
		4661	file descriptors again. Note that the replacement function has to close
		4662	the underlying OS handle.
		4663
		4664	=item EV_USE_WSASOCKET
		4665
		4666	If defined to be C<1>, libev will use C<WSASocket> to create its internal
		4667	communication socket, which works better in some environments. Otherwise,
		4668	the normal C<socket> function will be used, which works better in other
		4669	environments.
		4670
		4671	=item EV_USE_POLL
		4672
		4673	If defined to be C<1>, libev will compile in support for the C<poll>(2)
		4674	backend. Otherwise it will be enabled on non-win32 platforms. It
		4675	takes precedence over select.
		4676
		4677	=item EV_USE_EPOLL
		4678
		4679	If defined to be C<1>, libev will compile in support for the Linux
		4680	C<epoll>(7) backend. Its availability will be detected at runtime,
		4681	otherwise another method will be used as fallback. This is the preferred
		4682	backend for GNU/Linux systems. If undefined, it will be enabled if the
		4683	headers indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		4684
		4685	=item EV_USE_LINUXAIO
		4686
		4687	If defined to be C<1>, libev will compile in support for the Linux
		4688	aio backend. Due to it's currenbt limitations it has to be requested
		4689	explicitly. If undefined, it will be enabled on linux, otherwise
		4690	disabled.
		4691
		4692	=item EV_USE_KQUEUE
		4693
		4694	If defined to be C<1>, libev will compile in support for the BSD style
		4695	C<kqueue>(2) backend. Its actual availability will be detected at runtime,
		4696	otherwise another method will be used as fallback. This is the preferred
		4697	backend for BSD and BSD-like systems, although on most BSDs kqueue only
		4698	supports some types of fds correctly (the only platform we found that
		4699	supports ptys for example was NetBSD), so kqueue might be compiled in, but
		4700	not be used unless explicitly requested. The best way to use it is to find
		4701	out whether kqueue supports your type of fd properly and use an embedded
		4702	kqueue loop.
		4703
		4704	=item EV_USE_PORT
		4705
		4706	If defined to be C<1>, libev will compile in support for the Solaris
		4707	10 port style backend. Its availability will be detected at runtime,
		4708	otherwise another method will be used as fallback. This is the preferred
		4709	backend for Solaris 10 systems.
		4710
		4711	=item EV_USE_DEVPOLL
		4712
		4713	Reserved for future expansion, works like the USE symbols above.
		4714
		4715	=item EV_USE_INOTIFY
		4716
		4717	If defined to be C<1>, libev will compile in support for the Linux inotify
		4718	interface to speed up C<ev_stat> watchers. Its actual availability will
		4719	be detected at runtime. If undefined, it will be enabled if the headers
		4720	indicate GNU/Linux + Glibc 2.4 or newer, otherwise disabled.
		4721
		4722	=item EV_NO_SMP
		4723
		4724	If defined to be C<1>, libev will assume that memory is always coherent
		4725	between threads, that is, threads can be used, but threads never run on
		4726	different cpus (or different cpu cores). This reduces dependencies
		4727	and makes libev faster.
		4728
		4729	=item EV_NO_THREADS
		4730
		4731	If defined to be C<1>, libev will assume that it will never be called from
		4732	different threads (that includes signal handlers), which is a stronger
		4733	assumption than C<EV_NO_SMP>, above. This reduces dependencies and makes
		4734	libev faster.
		4735
		4736	=item EV_ATOMIC_T
		4737
		4738	Libev requires an integer type (suitable for storing C<0> or C<1>) whose
		4739	access is atomic with respect to other threads or signal contexts. No
		4740	such type is easily found in the C language, so you can provide your own
		4741	type that you know is safe for your purposes. It is used both for signal
		4742	handler "locking" as well as for signal and thread safety in C<ev_async>
		4743	watchers.
		4744
		4745	In the absence of this define, libev will use C<sig_atomic_t volatile>
		4746	(from F<signal.h>), which is usually good enough on most platforms.
		4747
		4748	=item EV_H (h)
		4749
		4750	The name of the F<ev.h> header file used to include it. The default if
		4751	undefined is C<"ev.h"> in F<event.h>, F<ev.c> and F<ev++.h>. This can be
		4752	used to virtually rename the F<ev.h> header file in case of conflicts.
		4753
		4754	=item EV_CONFIG_H (h)
		4755
		4756	If C<EV_STANDALONE> isn't C<1>, this variable can be used to override
		4757	F<ev.c>'s idea of where to find the F<config.h> file, similarly to
		4758	C<EV_H>, above.
		4759
		4760	=item EV_EVENT_H (h)
		4761
		4762	Similarly to C<EV_H>, this macro can be used to override F<event.c>'s idea
		4763	of how the F<event.h> header can be found, the default is C<"event.h">.
		4764
		4765	=item EV_PROTOTYPES (h)
		4766
		4767	If defined to be C<0>, then F<ev.h> will not define any function
		4768	prototypes, but still define all the structs and other symbols. This is
		4769	occasionally useful if you want to provide your own wrapper functions
		4770	around libev functions.
		4771
		4772	=item EV_MULTIPLICITY
		4773
		4774	If undefined or defined to C<1>, then all event-loop-specific functions
		4775	will have the C<struct ev_loop *> as first argument, and you can create
		4776	additional independent event loops. Otherwise there will be no support
		4777	for multiple event loops and there is no first event loop pointer
		4778	argument. Instead, all functions act on the single default loop.
		4779
		4780	Note that C<EV_DEFAULT> and C<EV_DEFAULT_> will no longer provide a
		4781	default loop when multiplicity is switched off - you always have to
		4782	initialise the loop manually in this case.
		4783
		4784	=item EV_MINPRI
		4785
		4786	=item EV_MAXPRI
		4787
		4788	The range of allowed priorities. C<EV_MINPRI> must be smaller or equal to
		4789	C<EV_MAXPRI>, but otherwise there are no non-obvious limitations. You can
		4790	provide for more priorities by overriding those symbols (usually defined
		4791	to be C<-2> and C<2>, respectively).
		4792
		4793	When doing priority-based operations, libev usually has to linearly search
		4794	all the priorities, so having many of them (hundreds) uses a lot of space
		4795	and time, so using the defaults of five priorities (-2 .. +2) is usually
		4796	fine.
		4797
		4798	If your embedding application does not need any priorities, defining these
		4799	both to C<0> will save some memory and CPU.
		4800
		4801	=item EV_PERIODIC_ENABLE, EV_IDLE_ENABLE, EV_EMBED_ENABLE, EV_STAT_ENABLE,
		4802	EV_PREPARE_ENABLE, EV_CHECK_ENABLE, EV_FORK_ENABLE, EV_SIGNAL_ENABLE,
		4803	EV_ASYNC_ENABLE, EV_CHILD_ENABLE.
		4804
		4805	If undefined or defined to be C<1> (and the platform supports it), then
		4806	the respective watcher type is supported. If defined to be C<0>, then it
		4807	is not. Disabling watcher types mainly saves code size.
		4808
		4809	=item EV_FEATURES
		4810
		4811	If you need to shave off some kilobytes of code at the expense of some
		4812	speed (but with the full API), you can define this symbol to request
		4813	certain subsets of functionality. The default is to enable all features
		4814	that can be enabled on the platform.
		4815
		4816	A typical way to use this symbol is to define it to C<0> (or to a bitset
		4817	with some broad features you want) and then selectively re-enable
		4818	additional parts you want, for example if you want everything minimal,
		4819	but multiple event loop support, async and child watchers and the poll
		4820	backend, use this:
		4821
		4822	#define EV_FEATURES 0
		4823	#define EV_MULTIPLICITY 1
		4824	#define EV_USE_POLL 1
		4825	#define EV_CHILD_ENABLE 1
		4826	#define EV_ASYNC_ENABLE 1
		4827
		4828	The actual value is a bitset, it can be a combination of the following
		4829	values (by default, all of these are enabled):
		4830
		4831	=over 4
		4832
		4833	=item C<1> - faster/larger code
		4834
		4835	Use larger code to speed up some operations.
		4836
		4837	Currently this is used to override some inlining decisions (enlarging the
		4838	code size by roughly 30% on amd64).
		4839
		4840	When optimising for size, use of compiler flags such as C<-Os> with
		4841	gcc is recommended, as well as C<-DNDEBUG>, as libev contains a number of
		4842	assertions.
		4843
		4844	The default is off when C<__OPTIMIZE_SIZE__> is defined by your compiler
		4845	(e.g. gcc with C<-Os>).
		4846
		4847	=item C<2> - faster/larger data structures
		4848
		4849	Replaces the small 2-heap for timer management by a faster 4-heap, larger
		4850	hash table sizes and so on. This will usually further increase code size
		4851	and can additionally have an effect on the size of data structures at
		4852	runtime.
		4853
		4854	The default is off when C<__OPTIMIZE_SIZE__> is defined by your compiler
		4855	(e.g. gcc with C<-Os>).
		4856
		4857	=item C<4> - full API configuration
		4858
		4859	This enables priorities (sets C<EV_MAXPRI>=2 and C<EV_MINPRI>=-2), and
		4860	enables multiplicity (C<EV_MULTIPLICITY>=1).
		4861
		4862	=item C<8> - full API
		4863
		4864	This enables a lot of the "lesser used" API functions. See C<ev.h> for
		4865	details on which parts of the API are still available without this
		4866	feature, and do not complain if this subset changes over time.
		4867
		4868	=item C<16> - enable all optional watcher types
		4869
		4870	Enables all optional watcher types. If you want to selectively enable
		4871	only some watcher types other than I/O and timers (e.g. prepare,
		4872	embed, async, child...) you can enable them manually by defining
		4873	C<EV_watchertype_ENABLE> to C<1> instead.
		4874
		4875	=item C<32> - enable all backends
		4876
		4877	This enables all backends - without this feature, you need to enable at
		4878	least one backend manually (C<EV_USE_SELECT> is a good choice).
		4879
		4880	=item C<64> - enable OS-specific "helper" APIs
		4881
		4882	Enable inotify, eventfd, signalfd and similar OS-specific helper APIs by
		4883	default.
		4884
		4885	=back
		4886
		4887	Compiling with C<gcc -Os -DEV_STANDALONE -DEV_USE_EPOLL=1 -DEV_FEATURES=0>
		4888	reduces the compiled size of libev from 24.7Kb code/2.8Kb data to 6.5Kb
		4889	code/0.3Kb data on my GNU/Linux amd64 system, while still giving you I/O
		4890	watchers, timers and monotonic clock support.
		4891
		4892	With an intelligent-enough linker (gcc+binutils are intelligent enough
		4893	when you use C<-Wl,--gc-sections -ffunction-sections>) functions unused by
		4894	your program might be left out as well - a binary starting a timer and an
		4895	I/O watcher then might come out at only 5Kb.
		4896
		4897	=item EV_API_STATIC
		4898
		4899	If this symbol is defined (by default it is not), then all identifiers
		4900	will have static linkage. This means that libev will not export any
		4901	identifiers, and you cannot link against libev anymore. This can be useful
		4902	when you embed libev, only want to use libev functions in a single file,
		4903	and do not want its identifiers to be visible.
		4904
		4905	To use this, define C<EV_API_STATIC> and include F<ev.c> in the file that
		4906	wants to use libev.
		4907
		4908	This option only works when libev is compiled with a C compiler, as C++
		4909	doesn't support the required declaration syntax.
		4910
		4911	=item EV_AVOID_STDIO
		4912
		4913	If this is set to C<1> at compiletime, then libev will avoid using stdio
		4914	functions (printf, scanf, perror etc.). This will increase the code size
		4915	somewhat, but if your program doesn't otherwise depend on stdio and your
		4916	libc allows it, this avoids linking in the stdio library which is quite
		4917	big.
		4918
		4919	Note that error messages might become less precise when this option is
		4920	enabled.
		4921
		4922	=item EV_NSIG
		4923
		4924	The highest supported signal number, +1 (or, the number of
		4925	signals): Normally, libev tries to deduce the maximum number of signals
		4926	automatically, but sometimes this fails, in which case it can be
		4927	specified. Also, using a lower number than detected (C<32> should be
		4928	good for about any system in existence) can save some memory, as libev
		4929	statically allocates some 12-24 bytes per signal number.
		4930
		4931	=item EV_PID_HASHSIZE
		4932
		4933	C<ev_child> watchers use a small hash table to distribute workload by
		4934	pid. The default size is C<16> (or C<1> with C<EV_FEATURES> disabled),
		4935	usually more than enough. If you need to manage thousands of children you
		4936	might want to increase this value (I<must> be a power of two).
		4937
		4938	=item EV_INOTIFY_HASHSIZE
		4939
		4940	C<ev_stat> watchers use a small hash table to distribute workload by
		4941	inotify watch id. The default size is C<16> (or C<1> with C<EV_FEATURES>
		4942	disabled), usually more than enough. If you need to manage thousands of
		4943	C<ev_stat> watchers you might want to increase this value (I<must> be a
		4944	power of two).
		4945
		4946	=item EV_USE_4HEAP
		4947
		4948	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		4949	timer and periodics heaps, libev uses a 4-heap when this symbol is defined
		4950	to C<1>. The 4-heap uses more complicated (longer) code but has noticeably
		4951	faster performance with many (thousands) of watchers.
		4952
		4953	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4954	will be C<0>.
		4955
		4956	=item EV_HEAP_CACHE_AT
		4957
		4958	Heaps are not very cache-efficient. To improve the cache-efficiency of the
		4959	timer and periodics heaps, libev can cache the timestamp (I<at>) within
		4960	the heap structure (selected by defining C<EV_HEAP_CACHE_AT> to C<1>),
		4961	which uses 8-12 bytes more per watcher and a few hundred bytes more code,
		4962	but avoids random read accesses on heap changes. This improves performance
		4963	noticeably with many (hundreds) of watchers.
		4964
		4965	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4966	will be C<0>.
		4967
		4968	=item EV_VERIFY
		4969
		4970	Controls how much internal verification (see C<ev_verify ()>) will
		4971	be done: If set to C<0>, no internal verification code will be compiled
		4972	in. If set to C<1>, then verification code will be compiled in, but not
		4973	called. If set to C<2>, then the internal verification code will be
		4974	called once per loop, which can slow down libev. If set to C<3>, then the
		4975	verification code will be called very frequently, which will slow down
		4976	libev considerably.
		4977
		4978	The default is C<1>, unless C<EV_FEATURES> overrides it, in which case it
		4979	will be C<0>.
		4980
		4981	=item EV_COMMON
		4982
		4983	By default, all watchers have a C<void *data> member. By redefining
		4984	this macro to something else you can include more and other types of
		4985	members. You have to define it each time you include one of the files,
		4986	though, and it must be identical each time.
		4987
		4988	For example, the perl EV module uses something like this:
		4989
		4990	#define EV_COMMON \
		4991	SV *self; /* contains this struct */ \
		4992	SV *cb_sv, *fh /* note no trailing ";" */
		4993
		4994	=item EV_CB_DECLARE (type)
		4995
		4996	=item EV_CB_INVOKE (watcher, revents)
		4997
		4998	=item ev_set_cb (ev, cb)
		4999
		5000	Can be used to change the callback member declaration in each watcher,
		5001	and the way callbacks are invoked and set. Must expand to a struct member
		5002	definition and a statement, respectively. See the F<ev.h> header file for
		5003	their default definitions. One possible use for overriding these is to
		5004	avoid the C<struct ev_loop *> as first argument in all cases, or to use
		5005	method calls instead of plain function calls in C++.
		5006
		5007	=back
		5008
		5009	=head2 EXPORTED API SYMBOLS
		5010
		5011	If you need to re-export the API (e.g. via a DLL) and you need a list of
		5012	exported symbols, you can use the provided F<Symbol.*> files which list
		5013	all public symbols, one per line:
		5014
		5015	Symbols.ev for libev proper
		5016	Symbols.event for the libevent emulation
		5017
		5018	This can also be used to rename all public symbols to avoid clashes with
		5019	multiple versions of libev linked together (which is obviously bad in
		5020	itself, but sometimes it is inconvenient to avoid this).
		5021
		5022	A sed command like this will create wrapper C<#define>'s that you need to
		5023	include before including F<ev.h>:
		5024
		5025	<Symbols.ev sed -e "s/.*/#define & myprefix_&/" >wrap.h
		5026
		5027	This would create a file F<wrap.h> which essentially looks like this:
		5028
		5029	#define ev_backend myprefix_ev_backend
		5030	#define ev_check_start myprefix_ev_check_start
		5031	#define ev_check_stop myprefix_ev_check_stop
		5032	...
		5033
		5034	=head2 EXAMPLES
		5035
		5036	For a real-world example of a program the includes libev
		5037	verbatim, you can have a look at the EV perl module
		5038	(L<http://software.schmorp.de/pkg/EV.html>). It has the libev files in
		5039	the F<libev/> subdirectory and includes them in the F<EV/EVAPI.h> (public
		5040	interface) and F<EV.xs> (implementation) files. Only the F<EV.xs> file
		5041	will be compiled. It is pretty complex because it provides its own header
		5042	file.
		5043
		5044	The usage in rxvt-unicode is simpler. It has a F<ev_cpp.h> header file
		5045	that everybody includes and which overrides some configure choices:
		5046
		5047	#define EV_FEATURES 8
		5048	#define EV_USE_SELECT 1
		5049	#define EV_PREPARE_ENABLE 1
		5050	#define EV_IDLE_ENABLE 1
		5051	#define EV_SIGNAL_ENABLE 1
		5052	#define EV_CHILD_ENABLE 1
		5053	#define EV_USE_STDEXCEPT 0
		5054	#define EV_CONFIG_H <config.h>
		5055
		5056	#include "ev++.h"
		5057
		5058	And a F<ev_cpp.C> implementation file that contains libev proper and is compiled:
		5059
		5060	#include "ev_cpp.h"
		5061	#include "ev.c"
		5062
		5063	=head1 INTERACTION WITH OTHER PROGRAMS, LIBRARIES OR THE ENVIRONMENT
		5064
		5065	=head2 THREADS AND COROUTINES
		5066
		5067	=head3 THREADS
		5068
		5069	All libev functions are reentrant and thread-safe unless explicitly
		5070	documented otherwise, but libev implements no locking itself. This means
		5071	that you can use as many loops as you want in parallel, as long as there
		5072	are no concurrent calls into any libev function with the same loop
		5073	parameter (C<ev_default_*> calls have an implicit default loop parameter,
		5074	of course): libev guarantees that different event loops share no data
		5075	structures that need any locking.
		5076
		5077	Or to put it differently: calls with different loop parameters can be done
		5078	concurrently from multiple threads, calls with the same loop parameter
		5079	must be done serially (but can be done from different threads, as long as
		5080	only one thread ever is inside a call at any point in time, e.g. by using
		5081	a mutex per loop).
		5082
		5083	Specifically to support threads (and signal handlers), libev implements
		5084	so-called C<ev_async> watchers, which allow some limited form of
		5085	concurrency on the same event loop, namely waking it up "from the
		5086	outside".
		5087
		5088	If you want to know which design (one loop, locking, or multiple loops
		5089	without or something else still) is best for your problem, then I cannot
		5090	help you, but here is some generic advice:
		5091
		5092	=over 4
		5093
		5094	=item * most applications have a main thread: use the default libev loop
		5095	in that thread, or create a separate thread running only the default loop.
		5096
		5097	This helps integrating other libraries or software modules that use libev
		5098	themselves and don't care/know about threading.
		5099
		5100	=item * one loop per thread is usually a good model.
		5101
		5102	Doing this is almost never wrong, sometimes a better-performance model
		5103	exists, but it is always a good start.
		5104
		5105	=item * other models exist, such as the leader/follower pattern, where one
		5106	loop is handed through multiple threads in a kind of round-robin fashion.
		5107
		5108	Choosing a model is hard - look around, learn, know that usually you can do
		5109	better than you currently do :-)
		5110
		5111	=item * often you need to talk to some other thread which blocks in the
		5112	event loop.
		5113
		5114	C<ev_async> watchers can be used to wake them up from other threads safely
		5115	(or from signal contexts...).
		5116
		5117	An example use would be to communicate signals or other events that only
		5118	work in the default loop by registering the signal watcher with the
		5119	default loop and triggering an C<ev_async> watcher from the default loop
		5120	watcher callback into the event loop interested in the signal.
		5121
		5122	=back
		5123
		5124	See also L</THREAD LOCKING EXAMPLE>.
		5125
		5126	=head3 COROUTINES
		5127
		5128	Libev is very accommodating to coroutines ("cooperative threads"):
		5129	libev fully supports nesting calls to its functions from different
		5130	coroutines (e.g. you can call C<ev_run> on the same loop from two
		5131	different coroutines, and switch freely between both coroutines running
		5132	the loop, as long as you don't confuse yourself). The only exception is
		5133	that you must not do this from C<ev_periodic> reschedule callbacks.
		5134
		5135	Care has been taken to ensure that libev does not keep local state inside
		5136	C<ev_run>, and other calls do not usually allow for coroutine switches as
		5137	they do not call any callbacks.
		5138
		5139	=head2 COMPILER WARNINGS
		5140
		5141	Depending on your compiler and compiler settings, you might get no or a
		5142	lot of warnings when compiling libev code. Some people are apparently
		5143	scared by this.
		5144
		5145	However, these are unavoidable for many reasons. For one, each compiler
		5146	has different warnings, and each user has different tastes regarding
		5147	warning options. "Warn-free" code therefore cannot be a goal except when
		5148	targeting a specific compiler and compiler-version.
		5149
		5150	Another reason is that some compiler warnings require elaborate
		5151	workarounds, or other changes to the code that make it less clear and less
		5152	maintainable.
		5153
		5154	And of course, some compiler warnings are just plain stupid, or simply
		5155	wrong (because they don't actually warn about the condition their message
		5156	seems to warn about). For example, certain older gcc versions had some
		5157	warnings that resulted in an extreme number of false positives. These have
		5158	been fixed, but some people still insist on making code warn-free with
		5159	such buggy versions.
		5160
		5161	While libev is written to generate as few warnings as possible,
		5162	"warn-free" code is not a goal, and it is recommended not to build libev
		5163	with any compiler warnings enabled unless you are prepared to cope with
		5164	them (e.g. by ignoring them). Remember that warnings are just that:
		5165	warnings, not errors, or proof of bugs.
		5166
		5167
		5168	=head2 VALGRIND
		5169
		5170	Valgrind has a special section here because it is a popular tool that is
		5171	highly useful. Unfortunately, valgrind reports are very hard to interpret.
		5172
		5173	If you think you found a bug (memory leak, uninitialised data access etc.)
		5174	in libev, then check twice: If valgrind reports something like:
		5175
		5176	==2274== definitely lost: 0 bytes in 0 blocks.
		5177	==2274== possibly lost: 0 bytes in 0 blocks.
		5178	==2274== still reachable: 256 bytes in 1 blocks.
		5179
		5180	Then there is no memory leak, just as memory accounted to global variables
		5181	is not a memleak - the memory is still being referenced, and didn't leak.
		5182
		5183	Similarly, under some circumstances, valgrind might report kernel bugs
		5184	as if it were a bug in libev (e.g. in realloc or in the poll backend,
		5185	although an acceptable workaround has been found here), or it might be
		5186	confused.
		5187
		5188	Keep in mind that valgrind is a very good tool, but only a tool. Don't
		5189	make it into some kind of religion.
		5190
		5191	If you are unsure about something, feel free to contact the mailing list
		5192	with the full valgrind report and an explanation on why you think this
		5193	is a bug in libev (best check the archives, too :). However, don't be
		5194	annoyed when you get a brisk "this is no bug" answer and take the chance
		5195	of learning how to interpret valgrind properly.
		5196
		5197	If you need, for some reason, empty reports from valgrind for your project
		5198	I suggest using suppression lists.
		5199
		5200
		5201	=head1 PORTABILITY NOTES
		5202
		5203	=head2 GNU/LINUX 32 BIT LIMITATIONS
		5204
		5205	GNU/Linux is the only common platform that supports 64 bit file/large file
		5206	interfaces but I<disables> them by default.
		5207
		5208	That means that libev compiled in the default environment doesn't support
		5209	files larger than 2GiB or so, which mainly affects C<ev_stat> watchers.
		5210
		5211	Unfortunately, many programs try to work around this GNU/Linux issue
		5212	by enabling the large file API, which makes them incompatible with the
		5213	standard libev compiled for their system.
		5214
		5215	Likewise, libev cannot enable the large file API itself as this would
		5216	suddenly make it incompatible to the default compile time environment,
		5217	i.e. all programs not using special compile switches.
		5218
		5219	=head2 OS/X AND DARWIN BUGS
		5220
		5221	The whole thing is a bug if you ask me - basically any system interface
		5222	you touch is broken, whether it is locales, poll, kqueue or even the
		5223	OpenGL drivers.
		5224
		5225	=head3 C<kqueue> is buggy
		5226
		5227	The kqueue syscall is broken in all known versions - most versions support
		5228	only sockets, many support pipes.
		5229
		5230	Libev tries to work around this by not using C<kqueue> by default on this
		5231	rotten platform, but of course you can still ask for it when creating a
		5232	loop - embedding a socket-only kqueue loop into a select-based one is
		5233	probably going to work well.
		5234
		5235	=head3 C<poll> is buggy
		5236
		5237	Instead of fixing C<kqueue>, Apple replaced their (working) C<poll>
		5238	implementation by something calling C<kqueue> internally around the 10.5.6
		5239	release, so now C<kqueue> I<and> C<poll> are broken.
		5240
		5241	Libev tries to work around this by not using C<poll> by default on
		5242	this rotten platform, but of course you can still ask for it when creating
		5243	a loop.
		5244
		5245	=head3 C<select> is buggy
		5246
		5247	All that's left is C<select>, and of course Apple found a way to fuck this
		5248	one up as well: On OS/X, C<select> actively limits the number of file
		5249	descriptors you can pass in to 1024 - your program suddenly crashes when
		5250	you use more.
		5251
		5252	There is an undocumented "workaround" for this - defining
		5253	C<_DARWIN_UNLIMITED_SELECT>, which libev tries to use, so select I<should>
		5254	work on OS/X.
		5255
		5256	=head2 SOLARIS PROBLEMS AND WORKAROUNDS
		5257
		5258	=head3 C<errno> reentrancy
		5259
		5260	The default compile environment on Solaris is unfortunately so
		5261	thread-unsafe that you can't even use components/libraries compiled
		5262	without C<-D_REENTRANT> in a threaded program, which, of course, isn't
		5263	defined by default. A valid, if stupid, implementation choice.
		5264
		5265	If you want to use libev in threaded environments you have to make sure
		5266	it's compiled with C<_REENTRANT> defined.
		5267
		5268	=head3 Event port backend
		5269
		5270	The scalable event interface for Solaris is called "event
		5271	ports". Unfortunately, this mechanism is very buggy in all major
		5272	releases. If you run into high CPU usage, your program freezes or you get
		5273	a large number of spurious wakeups, make sure you have all the relevant
		5274	and latest kernel patches applied. No, I don't know which ones, but there
		5275	are multiple ones to apply, and afterwards, event ports actually work
		5276	great.
		5277
		5278	If you can't get it to work, you can try running the program by setting
		5279	the environment variable C<LIBEV_FLAGS=3> to only allow C<poll> and
		5280	C<select> backends.
		5281
		5282	=head2 AIX POLL BUG
		5283
		5284	AIX unfortunately has a broken C<poll.h> header. Libev works around
		5285	this by trying to avoid the poll backend altogether (i.e. it's not even
		5286	compiled in), which normally isn't a big problem as C<select> works fine
		5287	with large bitsets on AIX, and AIX is dead anyway.
		5288
		5289	=head2 WIN32 PLATFORM LIMITATIONS AND WORKAROUNDS
		5290
		5291	=head3 General issues
		5292
		5293	Win32 doesn't support any of the standards (e.g. POSIX) that libev
		5294	requires, and its I/O model is fundamentally incompatible with the POSIX
		5295	model. Libev still offers limited functionality on this platform in
		5296	the form of the C<EVBACKEND_SELECT> backend, and only supports socket
		5297	descriptors. This only applies when using Win32 natively, not when using
		5298	e.g. cygwin. Actually, it only applies to the microsofts own compilers,
		5299	as every compiler comes with a slightly differently broken/incompatible
		5300	environment.
		5301
		5302	Lifting these limitations would basically require the full
		5303	re-implementation of the I/O system. If you are into this kind of thing,
		5304	then note that glib does exactly that for you in a very portable way (note
		5305	also that glib is the slowest event library known to man).
		5306
		5307	There is no supported compilation method available on windows except
		5308	embedding it into other applications.
		5309
		5310	Sensible signal handling is officially unsupported by Microsoft - libev
		5311	tries its best, but under most conditions, signals will simply not work.
		5312
		5313	Not a libev limitation but worth mentioning: windows apparently doesn't
		5314	accept large writes: instead of resulting in a partial write, windows will
		5315	either accept everything or return C<ENOBUFS> if the buffer is too large,
		5316	so make sure you only write small amounts into your sockets (less than a
		5317	megabyte seems safe, but this apparently depends on the amount of memory
		5318	available).
		5319
		5320	Due to the many, low, and arbitrary limits on the win32 platform and
		5321	the abysmal performance of winsockets, using a large number of sockets
		5322	is not recommended (and not reasonable). If your program needs to use
		5323	more than a hundred or so sockets, then likely it needs to use a totally
		5324	different implementation for windows, as libev offers the POSIX readiness
		5325	notification model, which cannot be implemented efficiently on windows
		5326	(due to Microsoft monopoly games).
		5327
		5328	A typical way to use libev under windows is to embed it (see the embedding
		5329	section for details) and use the following F<evwrap.h> header file instead
		5330	of F<ev.h>:
		5331
		5332	#define EV_STANDALONE /* keeps ev from requiring config.h */
		5333	#define EV_SELECT_IS_WINSOCKET 1 /* configure libev for windows select */
		5334
		5335	#include "ev.h"
		5336
		5337	And compile the following F<evwrap.c> file into your project (make sure
		5338	you do I<not> compile the F<ev.c> or any other embedded source files!):
		5339
		5340	#include "evwrap.h"
		5341	#include "ev.c"
		5342
		5343	=head3 The winsocket C<select> function
		5344
		5345	The winsocket C<select> function doesn't follow POSIX in that it
		5346	requires socket I<handles> and not socket I<file descriptors> (it is
		5347	also extremely buggy). This makes select very inefficient, and also
		5348	requires a mapping from file descriptors to socket handles (the Microsoft
		5349	C runtime provides the function C<_open_osfhandle> for this). See the
		5350	discussion of the C<EV_SELECT_USE_FD_SET>, C<EV_SELECT_IS_WINSOCKET> and
		5351	C<EV_FD_TO_WIN32_HANDLE> preprocessor symbols for more info.
		5352
		5353	The configuration for a "naked" win32 using the Microsoft runtime
		5354	libraries and raw winsocket select is:
		5355
		5356	#define EV_USE_SELECT 1
		5357	#define EV_SELECT_IS_WINSOCKET 1 /* forces EV_SELECT_USE_FD_SET, too */
		5358
		5359	Note that winsockets handling of fd sets is O(n), so you can easily get a
		5360	complexity in the O(n²) range when using win32.
		5361
		5362	=head3 Limited number of file descriptors
		5363
		5364	Windows has numerous arbitrary (and low) limits on things.
		5365
		5366	Early versions of winsocket's select only supported waiting for a maximum
		5367	of C<64> handles (probably owning to the fact that all windows kernels
		5368	can only wait for C<64> things at the same time internally; Microsoft
		5369	recommends spawning a chain of threads and wait for 63 handles and the
		5370	previous thread in each. Sounds great!).
		5371
		5372	Newer versions support more handles, but you need to define C<FD_SETSIZE>
		5373	to some high number (e.g. C<2048>) before compiling the winsocket select
		5374	call (which might be in libev or elsewhere, for example, perl and many
		5375	other interpreters do their own select emulation on windows).
		5376
		5377	Another limit is the number of file descriptors in the Microsoft runtime
		5378	libraries, which by default is C<64> (there must be a hidden I<64>
		5379	fetish or something like this inside Microsoft). You can increase this
		5380	by calling C<_setmaxstdio>, which can increase this limit to C<2048>
		5381	(another arbitrary limit), but is broken in many versions of the Microsoft
		5382	runtime libraries. This might get you to about C<512> or C<2048> sockets
		5383	(depending on windows version and/or the phase of the moon). To get more,
		5384	you need to wrap all I/O functions and provide your own fd management, but
		5385	the cost of calling select (O(n²)) will likely make this unworkable.
		5386
		5387	=head2 PORTABILITY REQUIREMENTS
		5388
		5389	In addition to a working ISO-C implementation and of course the
		5390	backend-specific APIs, libev relies on a few additional extensions:
		5391
		5392	=over 4
		5393
		5394	=item C<void (*)(ev_watcher_type *, int revents)> must have compatible
		5395	calling conventions regardless of C<ev_watcher_type *>.
		5396
		5397	Libev assumes not only that all watcher pointers have the same internal
		5398	structure (guaranteed by POSIX but not by ISO C for example), but it also
		5399	assumes that the same (machine) code can be used to call any watcher
		5400	callback: The watcher callbacks have different type signatures, but libev
		5401	calls them using an C<ev_watcher *> internally.
		5402
		5403	=item null pointers and integer zero are represented by 0 bytes
		5404
		5405	Libev uses C<memset> to initialise structs and arrays to C<0> bytes, and
		5406	relies on this setting pointers and integers to null.
		5407
		5408	=item pointer accesses must be thread-atomic
		5409
		5410	Accessing a pointer value must be atomic, it must both be readable and
		5411	writable in one piece - this is the case on all current architectures.
		5412
		5413	=item C<sig_atomic_t volatile> must be thread-atomic as well
		5414
		5415	The type C<sig_atomic_t volatile> (or whatever is defined as
		5416	C<EV_ATOMIC_T>) must be atomic with respect to accesses from different
		5417	threads. This is not part of the specification for C<sig_atomic_t>, but is
		5418	believed to be sufficiently portable.
		5419
		5420	=item C<sigprocmask> must work in a threaded environment
		5421
		5422	Libev uses C<sigprocmask> to temporarily block signals. This is not
		5423	allowed in a threaded program (C<pthread_sigmask> has to be used). Typical
		5424	pthread implementations will either allow C<sigprocmask> in the "main
		5425	thread" or will block signals process-wide, both behaviours would
		5426	be compatible with libev. Interaction between C<sigprocmask> and
		5427	C<pthread_sigmask> could complicate things, however.
		5428
		5429	The most portable way to handle signals is to block signals in all threads
		5430	except the initial one, and run the signal handling loop in the initial
		5431	thread as well.
		5432
		5433	=item C<long> must be large enough for common memory allocation sizes
		5434
		5435	To improve portability and simplify its API, libev uses C<long> internally
		5436	instead of C<size_t> when allocating its data structures. On non-POSIX
		5437	systems (Microsoft...) this might be unexpectedly low, but is still at
		5438	least 31 bits everywhere, which is enough for hundreds of millions of
		5439	watchers.
		5440
		5441	=item C<double> must hold a time value in seconds with enough accuracy
		5442
		5443	The type C<double> is used to represent timestamps. It is required to
		5444	have at least 51 bits of mantissa (and 9 bits of exponent), which is
		5445	good enough for at least into the year 4000 with millisecond accuracy
		5446	(the design goal for libev). This requirement is overfulfilled by
		5447	implementations using IEEE 754, which is basically all existing ones.
		5448
		5449	With IEEE 754 doubles, you get microsecond accuracy until at least the
		5450	year 2255 (and millisecond accuracy till the year 287396 - by then, libev
		5451	is either obsolete or somebody patched it to use C<long double> or
		5452	something like that, just kidding).
		5453
		5454	=back
		5455
		5456	If you know of other additional requirements drop me a note.
		5457
		5458
		5459	=head1 ALGORITHMIC COMPLEXITIES
		5460
		5461	In this section the complexities of (many of) the algorithms used inside
		5462	libev will be documented. For complexity discussions about backends see
		5463	the documentation for C<ev_default_init>.
		5464
		5465	All of the following are about amortised time: If an array needs to be
		5466	extended, libev needs to realloc and move the whole array, but this
		5467	happens asymptotically rarer with higher number of elements, so O(1) might
		5468	mean that libev does a lengthy realloc operation in rare cases, but on
		5469	average it is much faster and asymptotically approaches constant time.
		5470
		5471	=over 4
		5472
		5473	=item Starting and stopping timer/periodic watchers: O(log skipped_other_timers)
		5474
		5475	This means that, when you have a watcher that triggers in one hour and
		5476	there are 100 watchers that would trigger before that, then inserting will
		5477	have to skip roughly seven (C<ld 100>) of these watchers.
		5478
		5479	=item Changing timer/periodic watchers (by autorepeat or calling again): O(log skipped_other_timers)
		5480
		5481	That means that changing a timer costs less than removing/adding them,
		5482	as only the relative motion in the event queue has to be paid for.
		5483
		5484	=item Starting io/check/prepare/idle/signal/child/fork/async watchers: O(1)
		5485
		5486	These just add the watcher into an array or at the head of a list.
		5487
		5488	=item Stopping check/prepare/idle/fork/async watchers: O(1)
		5489
		5490	=item Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))
		5491
		5492	These watchers are stored in lists, so they need to be walked to find the
		5493	correct watcher to remove. The lists are usually short (you don't usually
		5494	have many watchers waiting for the same fd or signal: one is typical, two
		5495	is rare).
		5496
		5497	=item Finding the next timer in each loop iteration: O(1)
		5498
		5499	By virtue of using a binary or 4-heap, the next timer is always found at a
		5500	fixed position in the storage array.
		5501
		5502	=item Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)
		5503
		5504	A change means an I/O watcher gets started or stopped, which requires
		5505	libev to recalculate its status (and possibly tell the kernel, depending
		5506	on backend and whether C<ev_io_set> was used).
		5507
		5508	=item Activating one watcher (putting it into the pending state): O(1)
		5509
		5510	=item Priority handling: O(number_of_priorities)
		5511
		5512	Priorities are implemented by allocating some space for each
		5513	priority. When doing priority-based operations, libev usually has to
		5514	linearly search all the priorities, but starting/stopping and activating
		5515	watchers becomes O(1) with respect to priority handling.
		5516
		5517	=item Sending an ev_async: O(1)
		5518
		5519	=item Processing ev_async_send: O(number_of_async_watchers)
		5520
		5521	=item Processing signals: O(max_signal_number)
		5522
		5523	Sending involves a system call I<iff> there were no other C<ev_async_send>
		5524	calls in the current loop iteration and the loop is currently
		5525	blocked. Checking for async and signal events involves iterating over all
		5526	running async watchers or all signal numbers.
		5527
		5528	=back
		5529
		5530
		5531	=head1 PORTING FROM LIBEV 3.X TO 4.X
		5532
		5533	The major version 4 introduced some incompatible changes to the API.
		5534
		5535	At the moment, the C<ev.h> header file provides compatibility definitions
		5536	for all changes, so most programs should still compile. The compatibility
		5537	layer might be removed in later versions of libev, so better update to the
		5538	new API early than late.
		5539
		5540	=over 4
		5541
		5542	=item C<EV_COMPAT3> backwards compatibility mechanism
		5543
		5544	The backward compatibility mechanism can be controlled by
		5545	C<EV_COMPAT3>. See L</"PREPROCESSOR SYMBOLS/MACROS"> in the L</EMBEDDING>
		5546	section.
		5547
		5548	=item C<ev_default_destroy> and C<ev_default_fork> have been removed
		5549
		5550	These calls can be replaced easily by their C<ev_loop_xxx> counterparts:
		5551
		5552	ev_loop_destroy (EV_DEFAULT_UC);
		5553	ev_loop_fork (EV_DEFAULT);
		5554
		5555	=item function/symbol renames
		5556
		5557	A number of functions and symbols have been renamed:
		5558
		5559	ev_loop => ev_run
		5560	EVLOOP_NONBLOCK => EVRUN_NOWAIT
		5561	EVLOOP_ONESHOT => EVRUN_ONCE
		5562
		5563	ev_unloop => ev_break
		5564	EVUNLOOP_CANCEL => EVBREAK_CANCEL
		5565	EVUNLOOP_ONE => EVBREAK_ONE
		5566	EVUNLOOP_ALL => EVBREAK_ALL
		5567
		5568	EV_TIMEOUT => EV_TIMER
		5569
		5570	ev_loop_count => ev_iteration
		5571	ev_loop_depth => ev_depth
		5572	ev_loop_verify => ev_verify
		5573
		5574	Most functions working on C<struct ev_loop> objects don't have an
		5575	C<ev_loop_> prefix, so it was removed; C<ev_loop>, C<ev_unloop> and
		5576	associated constants have been renamed to not collide with the C<struct
		5577	ev_loop> anymore and C<EV_TIMER> now follows the same naming scheme
		5578	as all other watcher types. Note that C<ev_loop_fork> is still called
		5579	C<ev_loop_fork> because it would otherwise clash with the C<ev_fork>
		5580	typedef.
		5581
		5582	=item C<EV_MINIMAL> mechanism replaced by C<EV_FEATURES>
		5583
		5584	The preprocessor symbol C<EV_MINIMAL> has been replaced by a different
		5585	mechanism, C<EV_FEATURES>. Programs using C<EV_MINIMAL> usually compile
		5586	and work, but the library code will of course be larger.
		5587
		5588	=back
		5589
		5590
		5591	=head1 GLOSSARY
		5592
		5593	=over 4
		5594
		5595	=item active
		5596
		5597	A watcher is active as long as it has been started and not yet stopped.
		5598	See L</WATCHER STATES> for details.
		5599
		5600	=item application
		5601
		5602	In this document, an application is whatever is using libev.
		5603
		5604	=item backend
		5605
		5606	The part of the code dealing with the operating system interfaces.
		5607
		5608	=item callback
		5609
		5610	The address of a function that is called when some event has been
		5611	detected. Callbacks are being passed the event loop, the watcher that
		5612	received the event, and the actual event bitset.
		5613
		5614	=item callback/watcher invocation
		5615
		5616	The act of calling the callback associated with a watcher.
		5617
		5618	=item event
		5619
		5620	A change of state of some external event, such as data now being available
		5621	for reading on a file descriptor, time having passed or simply not having
		5622	any other events happening anymore.
		5623
		5624	In libev, events are represented as single bits (such as C<EV_READ> or
		5625	C<EV_TIMER>).
		5626
		5627	=item event library
		5628
		5629	A software package implementing an event model and loop.
		5630
		5631	=item event loop
		5632
		5633	An entity that handles and processes external events and converts them
		5634	into callback invocations.
		5635
		5636	=item event model
		5637
		5638	The model used to describe how an event loop handles and processes
		5639	watchers and events.
		5640
		5641	=item pending
		5642
		5643	A watcher is pending as soon as the corresponding event has been
		5644	detected. See L</WATCHER STATES> for details.
		5645
		5646	=item real time
		5647
		5648	The physical time that is observed. It is apparently strictly monotonic :)
		5649
		5650	=item wall-clock time
		5651
		5652	The time and date as shown on clocks. Unlike real time, it can actually
		5653	be wrong and jump forwards and backwards, e.g. when you adjust your
		5654	clock.
		5655
		5656	=item watcher
		5657
		5658	A data structure that describes interest in certain events. Watchers need
		5659	to be started (attached to an event loop) before they can receive events.
		5660
		5661	=back
825		5662
826	=head1 AUTHOR	5663	=head1 AUTHOR
827		5664
828	Marc Lehmann <libev@schmorp.de>.	5665	Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael
		5666	Magnusson and Emanuele Giaquinta, and minor corrections by many others.
829		5667

Diff Legend

–	Removed lines
+	Added lines
<	Changed lines
>	Changed lines